Coupled Electromagneti GabrielMendes

Coupled Electromagnetic and Thermal
Analysis of Electric Machines
Gabriel Dias Mendes
Supervised by:
Prof. Dr. Ângela Paula Barbosa de Silva Ferreira
Prof. Dr. Ednei Luiz Miotto
Bragança
2019 - 2020
Coupled Electromagnetic and Thermal
Analysis of Electric Machines
Gabriel Dias Mendes
Dissertation presented to the School of Technology and Management of

Polytechnic Institute of Bragança to the Fulfillment of the Requirements
for the Master of Science Degree in Industrial Engineering (Electrical
Engineering branch), in the scope of Double Degree with Federal
University of Technology - Paraná
Supervised by:
Prof. Dr. Ângela Paula Barbosa de Silva Ferreira
Prof. Dr. Ednei Luiz Miotto
Bragança
2019 - 2020
v
Abstract
The actual trend of the design process of electric machines is oriented to specific
requirements of the application and is no longer based in a standard structure. From this
point of view, the design procedure of electric machines became a multidisciplinary
process, involving electromagnetic, thermal, and mechanical modelling in a highly
iterative process between the different physics fields.
This dissertation deals with the design process of electric machines, proposing a
coupling methodology for the electromagnetic and thermal models which are
interrelated. The electromagnetic model establishes the main losses in electric machines:
iron and resistive losses. These losses are, in turn, the main heat sources, responsible for
heating and temperature distribution, i.e., the object of the thermal analysis, which affects
recursively the losses, due to parameter’ dependency on temperature. Also, the machine
temperature is crucial to maintain the lifetime of the machine. So, the coupled analysis is
mandatory to achieve the nowadays requirements of higher energy efficiency and power
density and cost reduction. Also, the coupled analysis enables optimization without the
need to build several prototypes, making this process more time and cost-efficiency.
Despite the temperature importance in electric machines, the thermal model was
overlooked over the years. However, it has been receiving more attention in the past
years.
In this work, the thermal modelling process is handled analytically and
numerically through finite element analysis (FEA), which is also used to obtain the
electromagnetic model. The modelling processes detailed during this work are applied
into a case study of a single-phase transformer with the rated power of 1 kW. The
numerical models were developed in the Ansys software suite, being the electromagnetic
model developed in Ansys Maxwell while the thermal model has developed in Ansys
Mechanical. At last, the coupling between the electromagnetic and thermal models was
accomplished in Ansys Workbench.
The results obtained from the models are compared and validated with the
experimental measurements of the losses and temperatures.
Keywords: Finite Element Analysis, Thermal Model, Electromagnetic Model, Coupled

Analysis, Electric Machine.
vii
Resumo
A tendência atual do processo de projeto de máquinas elétricas é orientada para
requisitos específicos de sua aplicação e não é mais baseada em uma estrutura padrão.
Deste ponto de vista, o procedimento de projeto de máquinas elétricas tornou-se um
processo multidisciplinar, envolvendo modelagem eletromagnética, térmica e mecânica
em um processo altamente iterativo entre os diferentes campos da física.
Esta dissertação trata do processo de projeto de máquinas elétricas, propondo uma
metodologia de acoplamento dos modelos eletromagnético e térmico que se inter-
relacionam. O modelo eletromagnético estabelece as principais perdas em máquinas
elétricas: perdas de ferro e resistivas. Essas perdas são, por sua vez, as principais fontes
de calor, responsáveis pelo aquecimento e distribuição de temperatura, ou seja, o objeto
da análise térmica, que afeta recursivamente as perdas, pois os parâmetros são
dependentes da temperatura. Além disso, a temperatura da máquina é crucial para
manter a vida útil da máquina. Assim, a análise acoplada é obrigatória para atender aos
requisitos atuais de maior eficiência energética e densidade de potência e redução de
custos.Além disso, a análise acoplada possibilita a otimização sem a necessidade de
construção de vários protótipos, tornando este processo mais eficiente em termos de
tempo e custos. Apesar da importância da temperatura nas máquinas elétricas, o modelo
térmico foi negligenciado ao longo dos anos. No entanto, tem recebido mais atenção nos
últimos anos.
Neste trabalho, o processo de modelagem térmica é tratado analiticamente e
numericamente por meio da análise de elementos finitos (FEA), que também é utilizada
para obter o modelo eletromagnético. Os processos de modelagem detalhados durante
este trabalho são aplicados em um estudo de caso de um transformador monofásico com
potência nominal de 1 kW. Os modelos numéricos foram desenvolvidos no pacote de
software Ansys, sendo o modelo eletromagnético desenvolvido no Ansys Maxwell
enquanto o modelo térmico foi desenvolvido no Ansys Mechanical. Por fim, o acoplamento
entre os modelos eletromagnético e térmico foi realizado no Ansys Workbench.
Os resultados obtidos com os modelos são comparados e validados com as
medições experimentais das perdas e temperaturas.
Palavras-chave: Análise por Elementos Finitos, Modelo Térmico, Modelo

Eletromagnético, Análise Acoplada, Máquinas Elétricas
ix
Acknowledgments
Firstly I want to thank God for the gift of life and salvation, enabling me to
accomplish this stage of my life. Without his grace and mercy, this work would not be
done.
Secondly, my endless gratitude for my parents, Lourival and Lourdes, and my
brother, Lucas, that helped in all stages of my life, and also believed in my potential.
This work is the result of the partnership of the Federal University of Technology
- Paraná (UTFPR) and the Polytechnic Institute of Bragança, therefore I’m thankful to be
part of two amazing institutions, that promoted my scientific development.
I want to acknowledge my supervisors PhD. Ângela Ferreira and PhD Ednei Miotto
for their support, mentoring, patience, comprehension and honest feedback. Without
their guidance, this work would not be accomplished.
Last but not least, I want to thank all my friends that supported me during this
stage of my life. In especial, my double degree colleagues, Victor, João, and Henrique, who
helped me during the period at Bragança.
xi
Table of Contents
List of Figures ........................................................................................................ xiii

List of Tables ............................................................................................................ xv
List of Abbreviations .......................................................................................... xvii
List of Symbols ....................................................................................................... xix
1 Introduction ........................................................................................................ 3
1.1 Motivations ....................................................................................................................................................... 3
1.2 Electric Machines Design and Modelling .............................................................................................. 4
1.3 Objectives........................................................................................................................................................... 5
1.4 Structure of the Work ................................................................................................................................... 6
2 Heat Generation and Transfer ...................................................................... 9
2.1 Heat Generation .............................................................................................................................................. 9
2.1.1 Resistive Loss ......................................................................................................................................... 9
2.1.2 Iron Loss ................................................................................................................................................ 11
2.2 Heat Transfer ................................................................................................................................................ 13
2.2.1 Conduction............................................................................................................................................ 13
2.2.2 Convection ............................................................................................................................................ 15
2.2.3 Radiation ............................................................................................................................................... 16
2.3 Summary ......................................................................................................................................................... 18
3 Analytical Thermal Modelling.....................................................................21
3.1 Thermal Resistance Network ................................................................................................................. 21
3.2 Thermal Resistance Network of Electric Machines ....................................................................... 22
3.2.1 Heat Generation.................................................................................................................................. 22
3.3 Thermal Resistances .................................................................................................................................. 23
3.3.1 Conduction............................................................................................................................................ 23
3.3.2 Convection ............................................................................................................................................ 24
3.3.3 Radiation ............................................................................................................................................... 26
3.3.4 Volume Element ................................................................................................................................. 26
3.4 Analytical Thermal Modelling of a Low Power Transformer .................................................... 27
3.4.1 Transformer Geometry.................................................................................................................... 27
3.4.2 Transformer Losses .......................................................................................................................... 28
3.4.3 Transformer Volume Element Modelling ................................................................................ 28
3.4.4 Winding Thermal Conductivity .................................................................................................... 32
xii
3.4.5 Results..................................................................................................................................................... 32
3.5 Summary ......................................................................................................................................................... 33
4 Numerical Modelling ...................................................................................... 37
4.1 Finite Element Method .............................................................................................................................. 38
4.2 Electromagnetic Finite Element Analysis .......................................................................................... 39
4.2.1 Numerical Electromagnetic Modelling of a Low Power Transformer.......................... 41
4.3 Thermal Finite Element Analysis .......................................................................................................... 46
4.3.1 Numerical Thermal Modelling of a Low Power Transformer .......................................... 47
4.4 Coupled Electromagnetic and Thermal Finite Element Analysis ............................................. 50
4.4.1 Numerical Coupled Modelling of a Low Power Transformer .......................................... 52
4.5 Summary ......................................................................................................................................................... 54
5 Results and Discussion .................................................................................. 57
5.1 Experimental Results ................................................................................................................................. 57
5.2 Summary of the Models Results and Discussion............................................................................. 59
5.2.1 Losses Results ...................................................................................................................................... 59
5.2.2 Temperature Results ........................................................................................................................ 59
5.3 Summary ......................................................................................................................................................... 63
6 Conclusion ......................................................................................................... 67
6.1 Future Work ................................................................................................................................................... 68
A. Transformer Characterization ................................................................... 81
B. Thermal Parameters and Formulation.................................................... 87
xiii
List of Figures
Figure 2.1 One-dimensional heat transfer by conduction (Adapted from [47]) .................................. 14
Figure 2.2 Convection between surface and air (Adapted from [48]). ..................................................... 15
Figure 2.3 Convection heat transfer types. (a) Forced convection. (b) Natural convection.
(Adapted from [47]). ..................................................................................................................................................... 16
Figure 2.4 Radiation heat transfer (Adapted from [48]). ............................................................................... 17
Figure 2.5 Directional distribution of emissivity (Adapted from [48]). ................................................... 18
Figure 3.1 Rectangle bar .............................................................................................................................................. 23
Figure 3.2 Hollow cylinder.......................................................................................................................................... 24
Figure 3.3 Convection surface ................................................................................................................................... 25
Figure 3.4 Volume element (Adapted from [57]) .............................................................................................. 26
Figure 3.5 Transformer geometry: (a) Symmetry characteristic (b) Modelled geometry ............... 27
Figure 3.6 Transformer volume elements: (a) core and (b) windings. .................................................... 29
Figure 3.7 Convection coefficients ........................................................................................................................... 30
Figure 3.8 Vc4 thermal resistances ........................................................................................................................... 31
Figure 3.9 Winding structure ..................................................................................................................................... 32
Figure 3.10 Transformer temperature spots ...................................................................................................... 33
Figure 4.1 Typical finite elements: (a) One-dimensional, (b) two-dimensional, (c) three-
dimensional. (Adapted from [61]). .......................................................................................................................... 38
Figure 4.2 Electromagnetic FEA flowchart. ......................................................................................................... 40
Figure 4.3 Transformer geometry. .......................................................................................................................... 42
Figure 4.4 Laminated steel properties: specific loss (a), B-H curve (b). .................................................. 42
Figure 4.5 Magnetization curve of a soft ferromagnetic material (Adapted from [71]) ................... 43
Figure 4.6 Transformer electromagnetic FEA: mesh. ...................................................................................... 44
Figure 4.7 Transformer electromagnetic FEA: magnetic flux density. ..................................................... 45
Figure 4.8 Transformer electromagnetic FEA: iron losses. ........................................................................... 45
Figure 4.9 Thermal FEA flowchart........................................................................................................................... 46
Figure 4.10 Transformer thermal FEA: Convection and radiation surfaces. ......................................... 48
Figure 4.11 Transformer thermal FEA: convection coefficient hc1. ........................................................... 48
Figure 4.12 Transformer thermal FEA: temperature distribution............................................................. 49
Figure 4.13 Transformer thermal FEA: heat flux. ............................................................................................. 50
Figure 4.14 Electromagnetic and thermal coupled analysis flowchart. ................................................... 52
Figure 4.15 Transformer coupled FEA: loss evolution.................................................................................... 53
Figure 4.16 Transformer coupled FEA: temperature evolution.................................................................. 53
Figure 5.1 Thermal image of the transformer after 200 minutes............................................................... 58
Figure 5.2 Transformer temperature spots ......................................................................................................... 60
Figure A.1 Single-phase transformer (a) and its nameplate (b).................................................................. 81
Figure A.2 Windings configuration.......................................................................................................................... 81
Figure A.3 Low voltage winding cross-section. .................................................................................................. 82
Figure A.4 Transformer core geometry parameters. ....................................................................................... 83
Figure B.1 Developed TRN considering the transformer's losses at ambient temperature ............ 90
Figure B.2 Developed TRN considering the transformer's losses at the steady-state temperature
................................................................................................................................................................................................ 92
xv
List of Tables
Table 2.1 Skin depth in copper. Adapted from [31] ......................................................................................... 10
Table 2.1 Typical values of the convection heat transfer [47] ..................................................................... 16
Table 3.1 Transformer losses .................................................................................................................................... 28
Table 3.2 TRN temperature results ......................................................................................................................... 33
Table 4.1 Power density distribution ..................................................................................................................... 47
Table 4.2 Thermal conductivities............................................................................................................................. 49
Table 5.1 Windings temperature measurement (IEC 60034). ..................................................................... 58
Table 5.2 Transformer losses: experimental and FEA. ................................................................................... 59
Table 5.3 Temperature results from the transformer thermal models ................................................... 60
Table 5.4 Temperature of the windings at steady-state ................................................................................. 61
Table 5.3 Trasformer temperature results: TRN, FEA, and coupled FEA ................................................ 62
Table A.1 Transformer tests data at room temperature. ............................................................................... 82
Table A.2 Transformer geometry parameters. ................................................................................................... 83
Table B.1 Materials thermal and physical properties. [4].............................................................................. 87
Table B.2 Windings thermal conductivities parameters. ............................................................................... 87
Table B.3 Convection coefficients relations. (Adapted from [48].) ............................................................ 88
Table B.4 Convection coefficients equations. ...................................................................................................... 89
Table B.5 TRN temperature results considering the transformer's losses at ambient temperature
................................................................................................................................................................................................ 91
Table B.6 TRN temperature results considering the transformer's losses at the steady-state
temperature ...................................................................................................................................................................... 93
xvii
List of Abbreviations
2D Two-dimensions
3D Three-dimensions
AC Alternating Current
CFD Computation Fluid Dynamics
DC Direct Current
FEA Finite Element Analysis
FEM Finite Element Method
LPTN Lumped Parameters Thermal Network
PMM Permanent Magnet Machine
PMSM Permanent Magnet Synchronous Motor
TRN Thermal Resistance Network
xix
List of Symbols
Temperature delta, °C or K
Temperature gradient, °C/m or K/m
A Surface area, m2
B Magnetic flux density, T
C0 Bertotti hysteresis loss coefficient
C1 Bertotti eddy current loss coefficient
C2 Bertotti excess eddy current loss coefficient
Cm Empirical parameter of Steinmetz’s equation
D Conductor diameter, m
F Frequency, Hz
h Convection heat transfer coefficient, W/m2/°C or W/m2/K
H Magnetic field strength, A/m
hrad Radiation heat transfer coefficient, W/m2/°C or W/m2/K
I Current, A
k Thermal conductivity, W/m/°C or W/m/K
kR Skin effect factor
L Length, m
nl Winding conductor layers
PBertotti Bertortti iron loss, W
PR Resistive loss, W
PSteinmetz Steinmetz iron loss, W
q Heat transfer rate, W
q’’ Heat flux, W/m2
R DC resistance, Ω
R0 DC resistance at temperature T0, Ω
RAC AC resistance, Ω
Rc Transformer’s core conduction thermal resistance, °C/W or K/W
Rch Transformer’s core convection thermal resistance, °C/W or K/W
xx
Rcond Conduction thermal resistance, °C/W or K/W

Rconv Convection thermal resistance, °C/W or K/W
Rcr Transformer’s core radiation thermal resistance, °C/W or K/W
Rrad Radiation thermal resistance, °C/W or K/W
Rw Transformer’s winding conduction thermal resistance, °C/W or K/W
Rwh Transformer’s winding convection thermal resistance, °C/W or K/W
Rwr Transformer’s winding radiation thermal resistance, °C/W or K/W
T Temperature, °C or K
T∞ Fluid temperature, °C or K
Ts Surface temperature, °C or K
Tsur Surrounding temperature, °C or K
V Voltage, V
α Resistivity temperature coefficient, or
β Steinmetz’s equation parameter

γ Steinmetz’s equation parameter
δ Skin depth, m
ε Emissivity
λ Reciprocal of the temperature coefficient of resistance at 0 °C of the conductor
material
σ Electric conductivity, S/m
σSB Stefan-Boltzmann constant, W/m2/K4
Chapter 1
Introduction
Motivations | 3
1 Introduction
1.1 Motivations
Electric machines are devices that are present in our daily activities, for instance,
refrigerators, washing machines, air conditioning systems, and also in industrial and
transportation applications. The aim to reduce the pollution caused by fossil fuel
increased the demand for more efficient devices and energy conversion systems.
According to the International Energy Agency [1], in 2018, the industry was the sector
responsible for consuming 41.95% of the world electricity. So, it is clear that there is a
global demand for electric machines to satisfy new applications requirements, namely
power density and energy efficiency, requiring a design methodology to support this
demand [2].
Electric machines are designed to fulfil the application’s requirements, e.g. torque,
power, speed, size. The classic design procedure is based on analytical methods to achieve
the electromagnetic requirements, while the other physics domains are handled with
empirical knowledge or even neglected [3]. As a consequence of the constraints and
requirements of the modern application, the design procedure became a multidisciplinary
process involving the relationship between different physics fields, e.g. electromagnetic,
thermodynamics, acoustics, structural mechanics. Thus, this work is motivated by the
necessity of handling the multiphysics nature of the electric machines in order to fulfil the
requirements of modern applications.
This work is focused on the thermal modelling of electric machines and its coupling
to the electromagnetic model. The thermal modelling of electric machines is crucial since
the device temperature is related to its performance and lifetime. For instance, if the
machine uses permanent magnets in its excitation system, for an operating temperature
of 100 °C, the magnetic flux may be reduced till 20% when compared with the magnetic
flux at room temperature [2], which impacts directly on the machine performance.
Besides that, the temperature rise increases the resistive loss of the machine reducing the
device efficiency. Additionally, there is the risk of demagnetization of permanent magnets
due to the temperature rise, and if a machine surpasses its insulation limit temperature
by 10 °C the insulation lifetime is cut by half [4]. Despite this strong relationship the
thermal modelling of electric machines has been neglected over the years especially for
small and medium-sized machines [5] but has received more attention in the past years,
4 | Introduction
in reference [6] it is possible to see the past years development to achieve better cooling
solutions of power traction machines.
1.2 Electric Machines Design and Modelling

There are two types of tools used to carry out the modelling of an electric machine:
analytical and numerical. Usually, the basic design and early dimensioning of the magnetic
and electric circuit are done by analytical equations, and numerical methods are used to
evaluate the device performance accurately [4]. Thus, both methods are used jointly to
exploit their advantages and mitigate their weakness. The use of numerical methods
became more popular with the advances on the digital computers, allowing an elevated
amount of complex and accurate calculations in a short period. Due to nowadays
computational resources, the numerical methods evolved, being able to simulate complex
physics phenomena with higher accuracy, although having the trade-off to be more time-
consuming along with requiring top tier computational power.
Due to the necessity to assimilate the multiphysics characteristic of electric
machines coupled analysis between two or more physics fields have been developed,
making the design process a high iterative process between the different physics analysis,
which allows better modelling and optimization of the device, without the drawbacks of
building different prototypes. Numerical methods are extensively used in coupled models
since multiphysics analysis is especially problematic from the analytical point of view.
Finite element analysis (FEA) is a popular numerical method that it is applicable to a
variety of physical phenomena, so it is highly used in multiphysics analysis. In [7], FEA is
used in the design process of a permanent magnet machine (PMM) covering the
electromagnetic, stress and thermal analysis of the device. Similarly, [8] focus on the
mechanical and thermal constraints to improve the rotor design of a PMM. Reference [9]
uses the acoustics and structural FEA coupling to reduce the noise of a switched
reluctance motor. Furthermore, in [10] the coupling analysis of electromagnetic,
mechanics and acoustic is used to investigate the faults on the pressure relief system in
the transformer’s tank caused by external short-circuit.
The scope of this work is the electromagnetic and thermal models of electric
machines. To accomplish the electromagnetic model analytically the reluctance network
may be used, which is a method typically used due to its simplicity, fast solution and
reasonable accuracy, that can also be applied to model complex magnet circuits [11]–[13].
Objectives | 5
Moreover, the numerical electromagnetic modelling is mostly conducted by FEA [4] which
has been used to explore design optimizations, estimate the losses in ferromagnetic
materials considering the non-uniform magnetic flux density distribution, and simulating
different working conditions [3], [14]–[17].
Concerning the thermal model, it can be achieved analytically through thermal
resistance network (TRN), which is also known as lumped parameters thermal network
(LPTN), and numerically by using FEA or computational fluid dynamics (CFD) [18]. TRN
is widely used [19]–[21] due to its simplicity along with a reasonable error since it is
basically analysed trough a dual electric circuit, allowing to use tools such as circuit
simulators to obtain the temperature results of the model. Therefore the method’s
challenge is to calculate the thermal resistances correctly, putting significant effort to
model the main heat paths [18]. However, the modelling of the thermal resistances of
complex parts can be very difficult and complicated to predict hot spots with enough
accuracy [22].
FEA is commonly used in thermal models [23]–[25] since it allows a massive
discretization of the geometry resulting in the precise temperature distribution of the
device, depending essentially on the accuracy of the thermal parameters used in the
model. The drawback of the FEA is that it relies on empirical relations to obtain the
convections coefficients, that it is also present in the TRN approach, so when the cooling
system of the machine involves complex fluid flow systems, the CFD takes place, being
used to simulate the fluid flow of the system and calculate its convection coefficient, which
was successfully used in [26]–[28]. Nonetheless, the CFD has the drawback of being
demanding on massive time and computational resources. Therefore hybrid approaches
combining CFD with FEA and TRN are usually used, like the one proposed in [22] to model
railway traction motors.
1.3 Objectives
Within the context set by the previous sections, the objectives of this work are
summarized in:
• Establish the main aspects of the heat generation and transfer on electric
machines that are necessary to understand the thermal model and its
characteristics, highlighting the relationship between the thermal and
electromagnetic analysis.
6 | Introduction
• Discuss the main analytical and numerical methods to achieve the thermal
model of electric machines, guiding through each modelling step.
• Develop a coupled electromagnetic and thermal analysis, to aid in the
design and derating process of electric machines.
• Apply the modelling methods discussed in this work in a case study, for
comparison and validation purposes.
1.4 Structure of the Work

Besides this chapter that presents the introduction of the work, covering the motivations,
objectives, and the state-of-the-art of the electric machines modelling process, the
dissertation’ form is structured as follows:
• Chapter 2: addresses the existing heat sources in electric machines and the
heat transfer mechanisms as well.
• Chapter 3: discusses analytical thermal modelling is and presents the
development process of three-dimensional TRN.
• Chapter 4: presents the electromagnetic and thermal numerical modelling
through FEA, and proposes the coupling methodology between the two
analyses.
• Chapter 5: the results obtained from the models developed for the work’s
case study are compared and confronted with experimental measures.
• Chapter 6: presents the main conclusions of the work, also draws the aims
for the future works related to this project.
Chapter 2
Heat Generation and Transfer
Heat Generation | 9
2 Heat Generation and Transfer

The power losses in electric machines are typically low when compared with the rated
power, although, they are capable of heating their components to temperatures where the
performance and lifetime are affected. Therefore, the device heat management is
necessary to grant its proper working conditions. This chapter addresses the principal
heat sources in electric machines and establishes the heat transfer background.
2.1 Heat Generation

Heat generation in electric machines occurs due to the losses present in the machine.
These losses can be separated into three categories, mechanical loss, electromagnetic loss,
and additional losses. Mechanical loss is present in rotating machines and relates with
ventilation and bearings friction. Electromagnetic losses are the losses originated by the
Joule heating in the windings, i.e., resistive loss, and the loss in the magnetic circuit, iron
loss. The additional losses consist of the electromagnetic loss not included by the resistive
and iron losses. They are related to the non-uniform distribution of the current in the
windings and to iron loss produced by the distortion of the magnetic flux caused by the
load current, and the iron loss that occurs in the machine frame [29]. These losses are not
determined by mathematical models since they are difficult to model and calculate. Thus
IEC standards define the additional losses as a portion of the input power [4].
The purpose of this section is to discuss and model the resistive and iron losses
since they correspond to the majority of the power loss in electric machines.
2.1.1 Resistive Loss

Resistive loss is the main loss source of an electric machine, they occur due to the damping
forces into the conductor, causing collisions between the atoms of the conductor and the
electric charges from the electric current flowing through the conductor, dissipating
energy in the heat form. This phenomenon is expressed by Joule’s first law
(2.1)
where I is the rms value of the current flowing through the conductor, and R is the direct
current (DC) resistance of the conductor.
Conductors under alternating current (AC) excitation can suffer from the skin
effect phenomenon, which increases the resistance and consequently the resistive loss.
This phenomenon occur due to the limitation of how much an electromagnetic wave can
10 | Heat Generation and Transfer
penetrate in a conductor, limiting the current flow surface area. The skin depth, δ,
determines how much the wave can penetrate the conductor and it is given by
(2.2)
being f the frequency of the current, µ the permeability and σ the conductivity of the
conductor. As δ is frequency dependent, it is necessary to evaluate in the frequency range
of the machine and for the significant harmonic components of the current. Table 2.1
shows the frequency influence on skin depth of the copper with S/m, and
permeability equal to the free space. It is possible to see the impact of this phenomenon
at high frequencies.
Table 2.1 Skin depth in copper. Adapted from [30]
Frequency (Hz) 10 50 100 500 104 108 1010

Skin depth (mm) 20.8 8.6 6.6 2.99 0.66
Considering the skin effect, the resistive loss is given by

(2.3)
where RAC is the AC resistance of the conductor, which is obtained by multiplying the DC
resistance by the skin effect factor kR [4], being dependent of the geometry and material
of the winding. For a round wire it can be approximated by
(2.4)
being nl the number of conductors layers, and D the diameter of the conductor. For
frequencies of the fundamental component and significant harmonic components where
δ is equal or higher than the conductor width, the skin effect factor, kR, can be considered
equal to 1.
The resistive loss has a strong relationship with the temperature, as the
temperature increases the resistance also increases. This dependence is expressed by
(2.5)
where R0 is the value of the resistance at the temperature T0, α is the temperature
coefficient of resistivity, that for materials typically used in windings such as the copper
and aluminium is respectively and [4], and T is the
actual temperature of the conductor. Using equation (2.5) it is possible to see that a
Heat Generation | 11
temperature rise of 55 °C in a copper wire, will result in a 20.96% increase of the

resistance, thus increasing the resistive loss.
2.1.2 Iron Loss

Iron loss is related to the phenomenon of magnetization in the ferromagnetic materials
used to create the magnetic flux path. It occurs due to the eddy currents flow on the
ferromagnetic material, causing Joule heating [31], and by the hysteresis loss caused by
the rotation of the magnetic dipoles on the ferromagnetic material, causing friction
between the neighbouring domains, thus energy consumption [32]. Despite the origin of
the iron loss being well-known, mathematical modelling still a challenge nowadays. This
is because of the complexity of the phenomena involved, requiring simplifications and
assumptions, the physical phenomena occur at a microscopic level, and the loss
formulation are developed at a macroscopic level.
In practice, the iron loss estimation is accomplished by empirical models or
equations based on the physical phenomena [33]. The usual approach to model iron loss
is to separate it into two distinct components, hysteresis and eddy current losses. The
eddy current loss is calculated for a uniform magnetization, which does not occur in
practice, and as consequence the eddy current loss is much higher than the calculated.
Thus, to compensate this error, iron loss model considers a third portion to the iron loss,
known as excess eddy current loss or anomalous loss [34]. An example of the loss
separation method is the Bertotti model for sinusoidal magnetization [35],
(2.6)
where Bmax is the maximum value of the magnetic flux density, C0 and C2 are the
coefficients related to the hysteresis and excess eddy current losses, being obtained from
experimental measures, C1 is the eddy current loss coefficient that is obtained by
(2.7)
being σ the electric conductivity and d the thickness of the electrical steel sheet. From the
equation above, it is possible to see the relation between the eddy current loss and the
thickness of the iron sheet, enlighten the reason why eddy current loss is massive on solid
objects.
Bertotti model has the drawback of being suitable only for applications under
sinusoidal magnetization at a limited frequency range, which is not the case for most
electric machines. To cover this drawback, models considering the effect of typical non-
sinusoidal waveforms were developed, for instance [36], where the average and rms
characteristics of the supply voltage are used to consider different waveforms, obtaining
errors of 5% to 6% of the loss prediction on the transformers used, and for induction
motors the errors were between 10%-20% In [37] are used the peak voltage and duty
cycle of the waveform, which resulted in a very good agreement with the experimental
data available for low and high frequency applications.
Many material sciences specialist contradicted the loss separation approach [38].
Graham in [31] establishes that there is no physical distinction between the hysteresis
loss and eddy current loss, the iron loss occurs whenever the magnetization changes,
which is done by domain wall motion, generating eddy currents and consequently loss.
The Steinmetz model is a well-known empirical equation that does not separate the iron
loss into different components,
(2.8)
where Cm, γ and β are empirical parameters obtained by fitting experimental curves [39],
and f is the frequency of the magnetization. Since this model was developed more than a
century ago, it is applicable only for sinusoidal magnetization. To apply this model in
applications with non-sinusoidal magnetization, researchers developed modified
versions to obtain an accurate prediction for arbitrary waveform, as the ones proposed in
[38] and [39], where it is used an equivalent frequency to model the waveform impact on
the iron loss.
One of the difficulties of calculating the iron loss is the fact that magnetization is
not uniformly distributed and, consequently, neither the losses are. This characteristic
sheds light to the FEA approach [14], [15], [17], since the FEA computes the flux density
for all elements of the geometry, enabling to calculate the loss for each element, therefore
getting the loss distribution on the geometry.
The iron loss is strongly dependent on the material properties, their characteristics
can vary about 30 % between different batches, and even in the same batch [40], bringing
a level of uncertainty between the specification sheet of the manufacturer and the actual
material used to build the electric machine. Mechanical stress caused by the
manufacturing process of shrink fitting and operating conditions of the machine also
affects the iron losses, studies were developed aiming the prediction and mitigation of
this factor [41]–[43].
Heat Transfer | 13
Differently from the resistive loss, the iron loss decreases with the temperature
increase. As the resistivity of the material increases with the temperature rise the eddy
currents are reduced, and consequently the heat dissipation. Several models considering
the temperature influence in iron loss have been developed. For instance, in [39] the
temperature influence is added to the Steinmetz model and in [16], [44], [45] the
temperature dependency is included to the separation loss model. The experimental data
of the iron loss in [16] reported that for a variety of frequencies and current loads the
temperature rise from 69°C to 100°C resulted in a maximum reduction of 12% of the iron
loss. Although, the temperature can have more impact on the iron loss for situations
where the machine operates at high temperature levels. Thus, the loss prediction without
the temperature dependency will result in a slight overestimation of the loss value.
2.2 Heat Transfer

Heat transfer is energy in transit due to a temperature difference, meaning that heat
transfer must occur whenever there is a temperature difference[46]. This temperature
difference can be between different physics states of matter, as a kettle (solid) and the
water (liquid), and even without a material medium between them (ultraviolet radiation).
The heat transfer occurs by three different modes: conduction, convection, and radiation.
2.2.1 Conduction
Heat transfer by conduction is related to the interaction between particles of a substance.
The particles with more energy, that also have higher temperatures, will transfer energy
to the surrounding particles with less energy. By this way, the heat flow by conduction
will occur from the higher temperature parts to the lower temperature spots. Conduction
occurs in the three states of matter, but within the electric machine context the conduction
is restrained to the solid-state, where the energy transport occurs by free electrons and
lattice vibrations[47].
The heat transfer rate is given by the Fourier’s law of heat conduction
(2.9)
expressing that the heat flow q is given by three parameters: thermal conductivity k of the
material, area A normal to the heat flow, and temperature gradient of the geometry.
The thermal conductivity parameter shows how much the material allows the heat flow.
The heat flux can be obtained from the equation (2.9):
(2.10)
The minus sign means that the heat is transferred from the higher temperature to
the lower temperature. Using a 3D cartesian coordinated system (x, y, z) and considering
an isotropic material, whereas k is constant for all directions, the heat flux assumes the
following form:
(2.11)
For anisotropic materials that have different thermal conductivities for each
direction, equation (2.11) can be written as
(2.12)
Not all cases need to be treated as a 3D heat conduction problem, it is possible to

assume that there will not have heat transfer in a specific direction, due to negligible
temperature difference or due to the lower thermal conductivity value when compared to
the others directions.
Figure 2.1 One-dimensional heat transfer by conduction (Adapted from [46])
The heat flux in the x-direction, illustrated in Figure 2.1, under steady-state
condition is
(2.13)
where Lx is the path length that separates the temperatures T1 and T2. This steady-state
equation applies to the other directions as well.
Since the conduction is dependent on the material medium, it is necessary to
determine the thermal conductivity of the material used in the devices. Generally
speaking, good electric conductors are also good thermal conductors, e.g. copper thermal
Heat Transfer | 15
conductivity is 394 W/m/°C, and electric insulators are also thermal insulators, e.g. k for
a typical insulation system of an electric machine is 0.2 W/m/°C. It should be mentioned
that the thermal conductivity is temperature dependent: thermal conductivity decreases
with increasing temperature for thermal conductors and increases it for thermal
insulators [4].
2.2.2 Convection
Convection is the heat transfer between a higher temperature region and a region of
cooler temperature [47], that is based on the energy transfer due to diffusion and motion
of the fluid. In this section will be developed to the heat convection between a surface.
In Figure 2.2, a surface (hot block) is in contact with a fluid (air), the temperature
of the surface is higher than the fluid temperature , so the heat of the hot block will
be transferred to the fluid. First, the heat transfer will occur by conduction from the
surface to the air layer adjacent to the surface, then the convection will be removing the
heated air and replacing it with cooler air.
Figure 2.2 Convection between surface and air (Adapted from [47]).
The velocity of the air on the surface is 0 m/s, but as soon the air takes distance
from the surface the velocity of the air will increase, implying a velocity gradient.
Similarly, the air will have a temperature gradient, the air temperature will be higher near
the surface, a value close to , and it will decrease with the distancing of the surface.
There are two types of convection, natural and forced convection, being the
difference between them the origin of the fluid motion. Regarding natural convection, the
fluid motion is caused by buoyancy forces due to the density difference caused by
temperature difference. Forced convection is based on the fluid motion induced by
external forces, such as a fan and a pump. Figure 2.3 illustrates both types of convection.
Figure 2.3 Convection heat transfer types. (a) Forced convection. (b) Natural convection. (Adapted from
[46]).
Independently of the type of convection, the heat transfer rate by convection is

given by Newton’s law of cooling
(2.14)
where h is the convection heat transfer coefficient, As is the surface area, Ts the surface
temperature, and T∞ the fluid temperature. This is a simple equation, but the challenge of
convection is on determining h. This coefficient is dependent on many factors, such as the
surface geometry and roughness, the fluid properties e.g. density, dynamic viscosity,
specific heat, etc., and the type of fluid flow [47].
The convection heat transfer coefficient, h, is typically defined through empirical
relations or computational fluid dynamics simulations [48]. Table 2.1 presents the value
range of h for some convection processes, highlighting the difference of the cooling
performance between the forced and natural convection, and also between the gases and
liquids.
Table 2.1 Typical values of the convection heat transfer [46]
Process h [W/ m²/°C]

Free Convection
Gases 2-25
Liquid 50-1000
Forced Convection
Gases 25-250
Liquids 100-20,000
2.2.3 Radiation
Radiation is the energy emitted by matter in the form of electromagnetic waves [47], due
to changes in the electron configurations of the atoms or molecules, the wavelength of
which lies in the range from 0.1 to 100 µm [4]. One of the most important properties of
the radiation is that it does not require a material medium. The radiation can occur from
Heat Transfer | 17
gases, liquids, and solids, but for electric machines, the focus is on the radiation from
solids surfaces.
The thermal radiation is the radiation emitted by a surface due to its temperature,
and the maximum emissive power of a surface is given by the Stefan-Boltzmann law
(2.15)
where σSB is the Stefan-Boltzmann constant that is equal to W/m2/K4, As is
the surface area, and Ts is the surface temperature. This equation considers an ideal
surface emitting at the maximum rate, i.e., a blackbody.
Real surfaces cannot emit at the maximum rate, so the equation (2.15) becomes
(2.16)
where ε is the emissivity of the surface that is dependent on the material, which varies
from 0 to 1. It should be noted that, when ε = 1, the surface is a blackbody.
Figure 2.4 shows a typical situation, where larger surfaces surround a smaller
surface, e.g. a room surrounding a motor frame. For such cases, the net rate of radiation
heat transfer between the surface and its surrounding is
(2.17)
being Tsur the surrounding surface temperature [47].
Figure 2.4 Radiation heat transfer (Adapted from [47]).
The emissivity of a black body surface is constant for all directions, this means that
it is a diffuse emitter. For real surfaces, the emissivity is not constant for all directions as
illustrated in Figure 2.5. Since the ratio between the emissivity in the normal direction
and in different directions is at maximum 30% for conductors and 5% for non-conductors,
it is reasonable to considerate the surface emissivity for all directions as the emissivity in
the normal direction [46].
Figure 2.5 Directional distribution of emissivity (Adapted from [47]).
2.3 Summary
This chapter discussed the main aspects of heat generation and heat transfer regarding
electric machines, detailing the physics phenomena and formulation of the main losses
found in electric machines. Also, the fundamentals of heat transfer necessary to
understand the thermal modelling process were briefly introduced. All three modes of
heat transfer are important in the thermal management of electric machines. Still,
convection is the primary mechanism to remove the heat from the device to the ambient,
along with the radiation, which has the performance similar to the natural convection.
Conduction is important to provide an efficient thermal flow, granting that inner heat
flows to the surfaces performing convection and radiation.
Chapter 3
Analytical Thermal Modelling
Thermal Resistance Network | 21
3 Analytical Thermal Modelling

In this chapter will be discussed the analytical thermal modelling of electric machines by
using the thermal resistance network approach. The concepts of this method are
presented, and the thermal modelling of an electric machine is described. The topic under
discussion is developed by applying the thermal resistance network analysis to a single-
phase transformer.
3.1 Thermal Resistance Network

The thermal resistance network (TRN) is an analytical method to model and solve steady-
state heat transfer problems by using the analogy between the heat flow and electric
circuits. It consists of the discretization of the problem domain into thermal resistances,
therefore with more discretization and more detailing the accuracy of TRN will increase,
but it also increases the complexity of solving and modelling.
Electric flow or electric current is defined by the well know Ohm’s law
(3.1)
which establishes the current I as the ratio of the potential difference V, or voltage, and
the electric resistance R. Comparing the analysis of electric circuits to the heat transfer
problem, it is possible to establish the followings analogies. The electric current occurs
due to a potential difference applied to electric resistance, in a similar way the heat flow
occurs due to a temperature difference. The electric current follows the path defined by
the electric resistance, so the thermal resistance defines the path of the heat flow as well.
With this analogy, it is possible to solve a heat transfer problem by solving its dual
electric circuit, where the current is replaced by the heat transfer rate q, the voltage by
the temperature difference, and the electric resistance by the thermal resistance. As the
thermal circuit is replaced by its dual electric circuit, the well-consolidated tools to solve
electric circuits can be exploited to solve the thermal circuit.
The temperature output of a TRN can be expressed as a matrix
(3.2)
which is equal to the thermal resistances matrix multiplied by the heat transfer rate
matrix. Each node presents a temperature output, therefore, if more temperature outputs
are wanted, more resistances are necessary, resulting in more discretization of the
22 | Analytical Thermal Modelling
problem domain. To perform an appropriate TRN it is necessary to evaluate what level of

discretization is necessary, and what domain region requires more discretization and
detail level.
Heat transfer is a three-dimensional phenomenon, but for some cases, i.e.,
anisotropic materials, it is possible to neglect the heat transfer in one direction, facilitating
the TRN modelling. Additionally, because most thermal resistances are temperature
dependent, the TRN resolution requires an iterative solution process, to model accurately
the process.
Section 3.3 explains the process to obtain the thermal resistance for each heat
transfer mode.
3.2 Thermal Resistance Network of Electric Machines

The use of TRN in electric machines is very convenient since engineers are familiar with
complex electric circuits. This method has been used for complex and modern electric
motors, such as the ones at the Formula E [21]. The study presented in [20] developed a
thermal network model for industrial induction motors, that resulted in a 5 °C error
between the TRN and the measured temperatures of the tested motor. The authors of [19]
also exploit the entropy generation to develop a thermal network of an induction motor
used to drive an electronic submersible pump.
The work developed in [49] exploits a TRN of a vehicle claw-pole alternator, and
investigates the improvement given by calibrating the convection coefficients used in the
model. A TRN of a 10 MW wind turbine is modelled in [50] and the results are compared
to 2D and 3D finite-element analysis, presenting temperature difference around 10%.
Despite of its simplicity the TRN is being widely used, receiving improvements
according to the needs of the project. It presents the flexibility of making a simple network
to obtain a quick temperature output with a certain error, or to make a detailed and
precise network without the need of a high computational power. Its usage on modern
applications shows that this method is relevant even with nowadays powerful computers
and software.
3.2.1 Heat Generation

The losses in electrical machines were previously explained in chapter 2. These losses are
incorporated in the TRN as current sources, and the way those losses are distributed in
the TRN has a significant impact on the model. The losses can be modelled by a single
Thermal Resistances | 23
current source, or they can be split in the TRN in a way that is more suitable to the heat
generation that occurs in the machine.
A precise heat generation distribution is not easy, since the losses are not uniformly
distributed and is difficult to have the losses distribution within the geometry. However,
it is recommended to put some effort in the allocation process of the heat sources,
according to the available information about the losses distribution.
3.3 Thermal Resistances

The challenge is to determine the thermal resistance since the heat transfer occurs by
three different modes, naturally, each mode will have a different resistance formulation.
Establishing the resistance as the ratio of the driving potential and the
corresponding transfer rate [46], a generic thermal resistance will be
(3.3)
Where is the temperature difference, and the heat transfer rate.
3.3.1 Conduction
Heat transfer by conduction is responsible for the main heat flow within the machine
material medium. Its correct modelling is crucial because it models the heat transfer paths
of the machine, as stated by [18], which emphasizes the importance of the effort to model
accurately the heat-transfer paths.
Figure 3.1 Rectangle bar
By using equation the thermal resistance for conduction, or conduction resistance,

is defined as
(3.4)
where L is the length of the heat transfer path, k the thermal conductivity, and A the cross-
section area. So, the thermal resistance of the Figure 3.1 bar for the x-direction conduction
is
(3.5)
Having one resistance on the x-direction gives the temperature information of the
two-ends of the bar, so to obtain more temperature outputs, it is necessary to divide the
path length according to the desired number of outputs and the desired temperature spot.
The same logic is applied to the other directions.
Figure 3.2 Hollow cylinder
As several electric machines have a hollow cylindrical shape, for instance their
stator core, the thermal resistance for conduction in the radial direction is estimated
through
(3.6)
being the outer radius, the inner radius, and L is the length of the cylinder.
The conduction resistance of the winding is a complex challenge, due to its

heterogeneous compound characteristic, and not uniformly geometry. As the windings
are the main source of heat, and it is where the higher temperatures are spotted, several
research focused on the windings conduction resistance were done, as the ones presented
in [51]–[55].
3.3.2 Convection
From equation the thermal resistance for convection, or convection resistance, is
obtained
Thermal Resistances | 25
(3.7)
where h is the convection heat transfer, and A is the surface area in contact with the fluid.
The challenge of the convection resistance is to obtain h, since it is dependent on many
factors, such as the type of fluid motion, the fluid used to cool the system.
By dividing the machine geometry into basic geometries it is possible to use the
convection relations available in heat transfer references, e.g. [46], [47], [56]. The
following equation is an example of a relation available in [47],
expressing h for a vertical plane with free air convection, where is the temperature
difference, in Celsius, between the surface and the air, and L is the length of the plate.
So the convection resistance of the surface in Figure 3.3 is
(3.8)
Figure 3.3 Convection surface
Although the equation is very simple, it addresses the temperature dependency of

the convection coefficient, which leads to an iterative process to obtain the convection
resistance, variable with the temperature results.
For machines with complex convection cooling systems, e.g. water jacket, oil spray,
oil as cooling fluid, it is necessary to put more effort to obtain the coefficients since these
cooling systems are more complex and the physics involved is not straightforward.
3.3.3 Radiation
Similarly to the convection resistance, the thermal resistance for radiation, or radiation
resistance is
(3.9)
where A is the surface area in contact with the surrounding and hrad is the radiation heat
transfer coefficient, which is given by
(3.10)
being ε the surface emissivity, the Stefan-Boltzmann constant, Ts the surface
temperature and Tsur the surrounding temperature.

Although the surface emissivity is not the same for all directions, it is possible to
approximate the emissivity of the surface to its emissivity in the normal direction, since
the difference is rarely higher than 30% [46]. To increase the radiation performance is
possible to choose the paint that has high emissivity.
The heat transfer through radiation is significant in electric machines, on the inside
and out of the machine. The authors of [57] evaluated the thermal resistances for
radiation of an induction motor, concluding that thermal radiation outside and inside of
the machine is comparable to free convection, pointing out the importance of the thermal
radiation on systems without forced convection. The main radiation paths inside the
machine are between the wires inside the slot and the stator lamination, and between the
end winding and the external frame.
3.3.4 Volume Element
Figure 3.4 Volume element (Adapted from [56])

Analytical Thermal Modelling of a Low Power Transformer | 27
Obtaining a TRN of a 3D geometry can be easily developed recurring to the volume

element method. Figure 3.4 shows a generic volume element for a cartesian coordinate
system, the subscript identifies the resistance axis.
By dividing the geometry into volume elements, the process of modelling the
thermal resistances and the connections between the resistances becomes easier. Using
more volume elements will lead to more precise results since the problem will have more
discretization.
3.4 Analytical Thermal Modelling of a Low Power

Transformer
To detail the process of modelling an electric machine, the analytical thermal modelling
method presented in this chapter was applied to model a low power single-phase
transformer. This transformer was chosen as case study of this work because, being the
first approach of the author to model electric machines, it is a simple static electric
machine, not involving mechanical loss and other aspects present in rotational machines,
and due to its availability in the laboratory to perform experimental tests. Appendix A
contains the details and characterization of the used transformer.
3.4.1 Transformer Geometry

The first step of thermal modelling is to spot the symmetry characteristics of the
geometry. Exploiting the symmetry characteristics allows a higher level of discretization
and detail, without the drawbacks of handling with so many resistances.
Figure 3.5 Transformer geometry: (a) Symmetry characteristic (b) Modelled geometry
Figure 3.5 shows the transformer geometry and exploits its symmetry
characteristics, so it is possible to see the existence of two symmetry planes, meaning that
it is only necessary to model ¼ of the transformer.
The following step consists in this geometry modelling into volume elements, such
as the one proposed in Figure 3.4, and connecting the volumes according to the direction
or directions that the volumes have contact. Each volume will have the number of thermal
resistances according to the contact that they make with other volumes, and with the
ambient.
3.4.2 Transformer Losses

The transformer is a static machine, thus the transformer losses correspond only to the
resistive loss on the windings, and to the iron loss on the core. Appendix A details the
procedure to obtain the losses experimentally.
It was assumed that the heat generation in the transformer occurs uniformly in
each material type, i.e., core and windings, although iron and resistive losses are
dependent of the iron saturation levels and current density, respectively, which are not
uniform within the machine. This simplification is assumed due to the fact that a non-
uniform distribution would require more detailing and inputs from the electromagnetic
model. By this way, the heat generated in the core will be equally distributed to all core’s
volumes elements, and the same development applies to the windings. Therefore, beyond
the thermal resistances, each transformer’s volume element has a heat source, modelled
as a current source.
Table 3.1 Transformer losses
Source Full Loss (W) ¼ Loss (W) Elements Element Loss (W)
Core 13 3.25 18 0.1806
Low Voltage Winding 29.26 7.32 4 1.8290
High Voltage Winding 20.34 5.09 4 1.2713
Table 3.1 details the modelled heat distribution. As the TRN represents only a ¼ of
the transformer, the losses in the TRN corresponds to ¼ of the total losses. At last, the loss
value is divided by the number of elements, according to the geometry discretization
underuse.
3.4.3 Transformer Volume Element Modelling

The objective of this section is to explain the process of obtaining the volume elements
thermal resistances and how they are connected with the ambient, and other volumes.
The same modelling process was applied to all volume elements, so for the sake of
simplicity, the process of a single volume element will be detailed.
Figure 3.6 (a) shows the volume elements of the core. In order to allow a better
visualization, labels of the volumes behind the ones numbered were omitted, being 9 plus
the volume number in the front. For example, the volume behind Vc1 is Vc10. And Figure
3.6 (b) shows the windings volume elements, the element Vw6 and Vw8 are below Vw5 and
Vw7, respectively, therefore it is not possible to visualize these two elements.
An important aspect to point out about this model is the fact that it considers there
is no heat transfer between the modelled portion of the transformer and its symmetry
counterpart. For example, it is not considered the heat transfer between Vc2 and the
neighbour from the other side of the symmetry plan. The model assumes a symmetry
boundary condition, i.e., the temperature of Vc2 is equal to the one of neighbour volume,
which in fact translates in neglecting the conduction resistance between them. Due to the
small temperature difference between the volumes near the boundary regions this
simplification is admissible. The results obtained from the thermal FEA (chapter 4)
corroborates with this simplification applied to the analytical model.
Figure 3.6 Transformer volume elements: (a) core and (b) windings.
Figure 3.7 shows the surfaces that perform convection and its corresponding
convection heat transfer coefficient, for aesthetic reasons, the winding surface facing
down (plane xz) is omitted, its corresponding coefficient is hw3. As the surfaces for
radiation and convection are the same, the label pattern of the radiation heat transfer
coefficient follows the same numbering of the convection coefficient of Figure 3.7, only
adding the subscript r, e.g. , , whereas subscripts c and w correspond to core and
winding elements, respectively.

Figure 3.7 Convection coefficients
The volume element chosen to explain the process is Vc4 because it is in contact
with other volumes and air. So, it is necessary to model the thermal resistances of the
volume in all three directions, and the resistances related to the convection and radiation.
First, the focus will be in the heat flow through Vc4 occurring in the x-direction. The
heat transfer occurs from Vc3 to Vc4, and from Vc4 to the ambient. So, one conduction
resistance connects Vc3 to the middle point of Vc4, and a similar resistance connects Vc4 to
the convection and radiation resistances, which are connected to the temperature
ambient. Since the heat is flowing in the x-direction the surface plane related to the
radiation and convection is the plane yz. For the y and z-directions this same logic is
applied, resulting in the TRN displayed in Figure 3.8.
Using the generic dimensions of Figure 3.4, the conduction thermal resistances of
Vc4 are
(3.11)
(3.12)
(3.13)
being the element dimensions for each direction, and the thermal
conductivity of the core in the direction specified by the correspondent subscript.

And for the convection and radiation, the thermal resistances are.
(3.14)
(3.15)
(3.16)
(3.17)
(3.18)
(3.19)
The subscript of each resistance specifies the heat transfer mode, r and h for radiation,
and convection, respectively, and the surface plane involved. The coefficients of heat
transfer are according to Figure 3.7.
Figure 3.8 Vc4 thermal resistances
Figure 3.8 shows the TRN of the Vc4, which involves a considerable number of
elements, indicating a complex and hard to handle TRN of the modelled geometry, due to
the massive number of resistances and connections. The heat source is connected to the
middle point of the element, that for Vc4 corresponds to iron loss (Piron) divided by the 18
core’s elements.
3.4.4 Winding Thermal Conductivity

Thermal resistances of the windings were obtained through the model developed
in [4], as presented in Figure 3.9, which shows the winding structure and parameters. As
the winding is a heterogenous element, it is necessary to obtain an average thermal
conductivity. Commonly the electric conductor is a good heat conductor as well, whereas
the electric insulator it is not a good heat conductor, e.g. copper conductivity is 1500 times
higher than epoxy conductivity, therefore it is possible to consider the electric conductor
as a perfect heat conductor.
Figure 3.9 Winding structure (Adapted from [4])
The average conductivity for this type of winding is given by
(3.20)
where ki is the thermal conductivity of the insulator material and the other parameters
are introduced in Figure 3.9.
3.4.5 Results
With the thermal resistances and heat sources properly determined, the complete TRN
can be solved by finding the equations related to temperature and solving it or using
modern tools to solve electric circuits. In this work, the TRN solution was accomplished
by using a circuit simulator, Multisim [58]. The thermal resistances of each volume
element were calculated and connected accordingly to the contact of each element with
the ambient and other elements, and the heat current related to the loss of each element
was connected on the middle point of the volume element, as shown in Figure 3.8. The
temperature results were obtained by placing voltage probes in the nodes of interest.
Figure 3.10 shows the temperature spots presented in Table 3.2. Appendix B shows the
TRN implemented in Multisim and its results for each element.
Summary | 33
Figure 3.10 Transformer temperature spots
Table 3.2 TRN temperature results
Spot Temperature (°C)

102
81.8
91.5
70.9
The higher temperature is spotted in the windings, as would be expected since they
are the main heat source of the transformer. T2 is higher than T4 because it is closer to the
windings than T4. Regarding T3 value, when compared with T2 its higher value may be
explained by the thermal conductivity of the core laminated steel since it is significantly
lower in the z-direction, than the conductivity in x and y-direction. Making it difficult the
heat flow in the direction normal to the lamination, thus the heat flow in the core occurs
majority on the x and y-direction. Along with, the core surface on plane xz is smaller than
the core surface on plane xy, implying a lower heat transfer rate at plane xz.
3.5 Summary
In this chapter was discussed the main aspects of the thermal modelling by using thermal
resistances network. This method is widely used due to its simplicity and fast solution,
whereas the precision of the model relies on the effort spent into modelling details,
discretization, and acquisition of thermal coefficients.
It was developed a TRN of a single-phase transformer, and its modelling process
was detailed and explained. The results obtained from the TRN were presented and
discussed.
Chapter 4
Numerical Modelling
Summary | 37
4 Numerical Modelling
Numerical methods require high levels of computational power that was not easily
available decades ago. With the advance of the digital computation, numerical methods
became more feasible and popular. One of the most popular numerical methods is the
finite element method (FEM), which was introduced in the early 1940s on the structural
analysis field, and now it is established for all different analysis fields.
In this chapter it will be discussed the use of FEM in electric machines modelling,
with focus in the electromagnetic and thermal fields. First, the concepts of FEM are
introduced, and afterwards, the electromagnetic and thermal analyses are presented, and
finally the electromagnetic and thermal coupled analysis is addressed.
The electromagnetic model is needed to provide the solution of the
electromagnetic fields of the machine, which will determine its rated power, losses and
other crucial aspects of the device. In the context of this work, the electromagnetic model
is responsible to obtain the losses of the machine, that are determined by the magnetic
flux density in soft ferromagnetic materials, i.e. the iron loss, and through the density
current in the conductors, the resistive loss.
The thermal model provides the temperature output of the machine due to its
losses, being those the main input to its thermal model. Both the analytical and numerical
thermal model considers the uniform heat generation within each part of the machine, by
using a constant loss density. This assumption is a simplification since in practice, iron
loss varies with the magnetic flux density along the iron volume. The electromagnetic and
thermal coupled analysis overcome this limitation by using the electromagnetic model
losses output as an input to the thermal model, that corresponds to a non-uniform heat
generation accordingly to the magnetic flux density and current density distribution in
the machine parts. Besides that, the coupled analysis uses the temperature output of the
thermal model as input to the electromagnetic model, therefore considering the
temperature dependency of the electromagnetic parameters.
The numerical modelling approach discussed during this chapter will be applied
to the case study of a low power single-phase transformer, that is detailed and
characterized in Appendix A. Also, its analytical thermal model was developed in section
3.4.
38 | Numerical Modelling
4.1 Finite Element Method

Finite element method (FEM) is a numeric technique developed to solve engineer
problems related to physics phenomena described by differential equations. These
elements are connected via nodes where the unknowns are to be determined. Algebraic
approximations of the differential equations that describe the physics are applied to
determine the element solution. The finite element equations are assembled into a set of
algebraic equations, also the boundary conditions are imposed into this set of equations.
At last the set of equations is solved, providing the solution of each node [59].
Three factors are important to FEM, differential equations, boundary conditions,
and mesh geometry. Differential equations and boundary conditions are related to the
physic phenomena properties, such as the differential equation of the heat conduction and
the ambient temperature for a heat transfer problem. The geometry is the domain where
the physics phenomena are occurring, it is divided into finite elements, and these finite
elements can have different shapes (see Figure 4.1), a set of finite elements is called mesh.
So the meshing process needs to be done carefully to discretize the geometry into several
elements with appropriate shape, aiming the desirable precision with a reasonable
computational cost.
Figure 4.1 Typical finite elements: (a) One-dimensional, (b) two-dimensional, (c) three-dimensional. (Adapted
from [60]).
The FEM is capable to solve 2D and 3D problems, therefore it is necessary to choose

which geometry, 2D or 3D analysis, is more suitable for the model. The 2D geometry
should be used in situations where variables under analysis do not change in one
Electromagnetic Finite Element Analysis | 39
direction, for instance, the magnetic flux density variation can be ignored on the direction
normal to the lamination of a laminated electric steel. Also in situations where the heat
conduction in one direction can be ignored, and the convection and radiation of the
surfaces on that direction can be neglected, 2D analysis may be conducted. Thus, the 3D
geometry should be used in situations where all directions must be considered, which
increases the computational weight of the solution, but corresponds to a more precise
solution since the modelled electric machine corresponds to a 3D geometry.
FEM is implemented on a variety of software, allowing the exploitation of the tool
without diving into the mathematical formulation of the method, but rather focusing on
the knowledge of the physics phenomena, and being able to analyse correctly the FEM
solution outputs. Therefore, the objective of this work is not to develop a FEM but use the
FEM tools available on the market. Using FEM to study or analyse a phenomenon is called
Finite Element Analysis (FEA), this is a powerful tool to quickly solve physic phenomena
problems, helping engineers in several areas, such as the design process of a machine
without having the requirement of building various prototypes.
4.2 Electromagnetic Finite Element Analysis

Electromagnetic phenomena are described by Maxwell’s equations, being the
electromagnetic FEA based on their differential form. Several studies regarding the
electromagnetic model of electric machines were done using FEA, one of the main topics
of these researches is the loss prediction and its mitigation. 3D FEA simulation is
performed in [61] to study the losses and the electromagnetic field of a 90 VA transformer,
in a similar way, [62] studies the iron loss of a 25 kVA distribution transformer under non-
sinusoidal voltage, which is important due to the nowadays high usage of power
electronics with their non-sinusoidal current waveform, causing distortions on the
system supply. Reference [63] recommends the utilization of FEA to obtain accurate local
electromagnetic field information to calculate iron loss.
Electromagnetic FEA is also important to explore the design efficiency of electric
machines, [64] uses the FEA simulation to compare two design choices of a permanent
magnet synchronous motor (PMSM). Whereas [65] investigate the impact manufacturing
tolerance of permanent magnets on PMSM. It is also possible to analyse the
electromagnetic fields and losses of electric machines with faults, such as [66] does on
induction motors with broken bars. FEA is a powerful tool which gives a lot of information
on the electromagnetic field in the machine, allowing the user to explore a variety of
applications.
Figure 4.2 Electromagnetic FEA flowchart.
Figure 4.2 shows a flowchart of the generic setup of an electromagnetic FEA

simulation. The first step to solve any problem is to identify which outputs are needed
and which inputs are required. The geometry of an electric machine is 3D, thus for a more
accurate model the 3D geometry should be used, although 2D solutions are suitable for
situations where the physical phenomenon in one direction can be ignored.
Material data of the components are important to enhance the constitutive
relations with accuracy, which may be complicated especially when the materials are
nonlinear, anisotropic and inhomogeneous, like the iron core of electrical machines. The
minimum data required for soft ferromagnetic materials is the B-H and specific loss
curves, while for materials used in electrical circuits, the electric conductivity is
mandatory.
To improve the performance of the FEM different solvers were developed, which
were obtained by defining some of the field characteristics, e.g. magnetostatic solver the
magnetic field is only produced by constant current flow or permanent magnets, this field
characteristic specification allows a more efficient solver. As each solver is developed to
fulfil a purpose it is important to read the software documentation to choose the solver
type that suits the user application.
The meshing step is where the geometry is divided into finite elements, and it is
necessary to pay attention to this step, a more refined mesh will have more elements,
hence more calculations and more time to obtain the solution. FEA software assists this
process by informing mesh metrics, allowing the user to evaluate if the mesh gives the
desired level of discretization.
Excitations are chosen according to the solver type and application, they can be DC
sources, permanent magnets or time-dependent sources. In more advanced software
packages it is possible to use power electronics allowing to simulate complex working
conditions.
The boundary conditions are defined accordingly to the physics phenomena, and
the characteristics of the device, the appropriate use of the boundary conditions is
necessary to model the physics phenomena along the domain boundary correctly. Thus,
each solver type make available the boundary conditions accordingly to the physics
phenomena. Commonly, are used two boundaries conditions, the Neumann conditions
where it defines the value of the solution derivative on the domain boundary, and the
Dirichlet conditions, which specifies the solution value along the domain boundary.
If the steps above were properly done, now it is possible to run the FEM and
analyse its solution. Now begins the analysis step, where the user needs to use his
phenomena knowledge to evaluate the FEM results. Without the proper knowledge of the
physical phenomena, the FEM solution is just a set of numbers, thus the final task of the
user is to transform this set of numbers into tangible information.
4.2.1 Numerical Electromagnetic Modelling of a Low Power

Transformer
In the context of this work, the electromagnetic model is developed to obtain the
electromagnetic losses of the machine and their non-uniform distribution. Therefore, the
electromagnetic model of the transformer was developed to match the rated working
conditions, hence the resistive and iron losses under the rated working conditions. The
electromagnetic FEA was developed by using the Ansys Maxwell 2019 R3 [67] software
package. Appendix A presents more details about the transformer.
The interest of this work is to establish the relation between heat generation and
its management in electric machines. Therefore the output wanted is the losses
distribution on the transformer, so it is necessary to use a 3D geometry (Figure 4.3) to
obtain the distribution of the iron loss on the core’s volume and the resistive loss in the
windings’ volume.
Figure 4.3 Transformer geometry.
The material data required to model the transformer are the core’s B-H curve and
specific loss, and the electric conductivity of the windings. Figure 4.4 presents the core’s
material data, that was adapted from one manufacturer [68]. The windings are made of
round wire impregnated copper, which has the conductivity of S/m [4].
Figure 4.4 Laminated steel properties: specific loss (a), B-H curve (b).
Next step is to choose the solver type: the magnetostatic solver is capable to
calculate only the resistive loss; to obtain both losses, resistive and iron losses, the
transient solver is required in order to simulate the magnetic flux variation[69]. Since it
is a transient simulation, it is necessary to specify the time step and time of the simulation.
The desired output is the losses, so the simulation needs to represent at least one
complete period of the excitation cycle, which, according with the rated characteristics of
the transformer, is 50 Hz or 20 ms.
The iron loss is obtained by the Ansys Maxwell built-in iron loss calculator that is
based on the Bertotti model [35], that was addressed in section 2.1.2. The coefficients
used in the iron loss model are automatically determined by the software, after providing
the specific loss curve (Figure 4.4 (a)), the electric conductivity of the electrical steel sheet,
and its thickness, that was defined as S m and 0.5 mm, respectively.
Figure 4.5 shows the magnetization curve of a soft ferromagnetic material, that
expresses the relation between B and H. In the point a the magnetic domains are randomly
oriented which results in a zero magnetization. As the external magnetic field increases
the magnetic domains will follow the orientation of the field, increasing magnetic flux
density. The path a – b is called the initial magnetization curve. As H decreases B will also
decrease, but it will not follow the initial curve (b – a), it will follow the path b – c. This
characteristic of not following the initial path of the magnetization is called hysteresis,
and the magnetization path of the ferromagnetic material after the first magnetization
(b – c – d – e – b), is called hysteresis loop.
Figure 4.5 Magnetization curve of a soft ferromagnetic material (Adapted from [70])
As the hysteresis loop is only established after the initial magnetization curve, the
iron loss must be considered after the initial magnetization curve, which occurs in the first
half period of the excitation. For the excitation applied to the transformer the time taken
to occur the first magnetization is 10 ms, thus the simulation must cover at least 10 ms
plus one complete period of the excitation. So, the time defined for the solver is 30 ms
with a time step of 0.5 ms.
Figure 4.6 Transformer electromagnetic FEA: mesh.
Figure 4.6 shows the transformer’s mesh, since the 3D transients simulation is
demanding on computational resources the mesh size was adjusted by the software
parameters, conciliating the solution precision and computational cost.
The excitations of the transformer were done by firstly setting up the number of
turns of each winding and its current flowing direction since it is a shell type transformer,
the directions of the windings are opposite. The electromagnetic models the rated
working conditions of the transformer, because in these conditions the nominal current
is flowing through the windings and the magnetic flux density on the core corresponds to
the rated voltage, granting that both losses are addressed in their rated values. To
simulate the rated working conditions, an external circuit excites the high voltage winding
with its rated voltage and connects a resistive load to the low voltage winding which
imposes the rated currents to the windings.
It was applied an insulating boundary condition between the windings, which
models the insulation material between them, this condition defines that the current
cannot cross the boundary, H is tangential to the boundary and the flux cannot cross it
[69]. Since the interest of the analysis is in the electromagnetic field of the transformer, it
was applied a Naumann boundary condition on the transformer’s external boundaries,
where it defines the H is tangential and the flux cannot cross the boundary. Thus, it does
not calculate the magnetic field on the ambient around the transformer, isolating the
transformer from the ambient.
Figure 4.7 Transformer electromagnetic FEA: magnetic flux density.
With all the setup properly done it is possible to run the solver and analyse the
solution. Figure 4.7 shows the magnetic field flux density in the transformer core at 15 ms,
the maximum value is 1.30 T, which is far from the saturation level of the ferromagnetic
material, and the high values occurs only in the edges close to windings.
Figure 4.8 Transformer electromagnetic FEA: iron loss.
Figure 4.8 displays the iron loss of the core during the simulation, is possible to see
the difference on the iron loss during the first magnetization curve, and after the
hysteresis loop establishment, the average iron loss between 10 ms and 30 ms is 13.78 W.
Since, ANSYS Maxwell iron loss calculation is based on the loss separation approach, is
possible to observe the contribution of each portion of iron losses (hysteresis, eddy-
current, excess eddy-current) to the total iron loss. And the average resistive loss of the
windings at ambient temperature is 44.12 W.
4.3 Thermal Finite Element Analysis

The heat transfer phenomena used in thermal FEA were previously described in chapter
2. FEA allows high levels of discretization granting accurate solutions regarding the heat
transfer through conduction, which provides an accurate temperature distribution in the
machine’s geometry. Reference [71] exploits this FEA advantage to investigate local
temperatures in the tank of a power transformer, in a similar way [24] studies the thermal
impact on the bushings of a transformer. The determination of the convection heat
transfer coefficient is based on the same relations used in the TRN modelling, although in
TRN the coefficient is considered as a constant value, whereas in FEA the implementation
of the temperature dependency of the convection and radiation coefficient is
straightforward. Figure 4.9 shows a flowchart of the thermal FEA model.
Figure 4.9 Thermal FEA flowchart.
Typically the thermal steady-state of the machine is the required state, because it
reflects the stable temperatures on working conditions, though, the transient solution is
important to see how long the machine takes to heat up and cools down, which can be
crucial for applications that require several starts. Reference [25] is an example where
the thermal transient solution is used to investigate the temperatures of an induction
motor during the starting process.
Thermal Finite Element Analysis | 47
Since the losses are generated inside of the machines they are treated as internal
heat generation, which are modelled by a power density (W/m3) that can be distributed
uniformly within the geometry, or in a way that better resembles the real heat generation.
Next step is to determine which surfaces will perform convection and their respective
convection coefficients. For the radiation is the same process as the convection, with the
difference that instead of convection coefficients, is necessary to input the emissivity of
the surface. The thermal conductivity of the machine materials will define the heat
conduction, for a anisotropic material is necessary to input the thermal conductivity of
each direction. Finally, with the heat distribution defined and the heat transfer
mechanisms characterized, it is possible to run the thermal FEM solution and procedure
its analysis.
4.3.1 Numerical Thermal Modelling of a Low Power Transformer

The thermal FEA simulation of the transformer characterized in Appendix A was
performed, by considering an uniform distribution of the losses on their respective parts.
Ansys Mechanical 2019 R3 [72] software package was used to perform the thermal FEA
model.
The 3D geometry is the same used in the electromagnetic FEA (see Figure 4.3).
Since the desired output is the steady-state temperature distribution of the transformer,
it is possible to have a more refined mesh without having a significant impact on the
solution time. Table 4.1 shows the power density values applied to the model.
Table 4.1 Power density distribution
Geometry Loss (W) Power density (W/m3)

Core
Low voltage winding
High voltage winding
Figure 4.10 shows the surfaces that will perform convection and radiation heat
transfer with the corresponding coefficient. As the surfaces for radiation and convection
are the same, the label pattern of the radiation heat transfer coefficient follows the same
numbering of the convection coefficient, only the subscript r is added, e.g. hrc1, hrw2,
whereas subscripts c and w correspond to core and winding surfaces, respectively. For
aesthetic reasons, the winding surface facing down (plane xz) is omitted, its
corresponding coefficient is hw3.
In contrast to the TRN, where the convection coefficient is considered a constant

value, the convection coefficients are inserted in Ansys Mechanical as a temperature
surface function, calculating more accurately the convection heat transfer. The convection
coefficients were defined by the same empirical relations used in the analytical model.
Figure 4.11 shows one of the coefficients used. The temperature also impacts in the
thermal conductivity of the materials, however for most cases it can be used a constant
average thermal conductivity. In this work the thermal conductivity is considered
constant, due to the unavailability of the parameters necessary to consider this
temperature dependency, and the thermal conductivity of resembling materials presents
slightly variation for the temperature rise occurring on the transformer.
Figure 4.10 Transformer thermal FEA: Convection and radiation surfaces.
Figure 4.11 Transformer thermal FEA: convection coefficient hc1.
The windings surfaces that are wrapped in a black electrical tape has the emissivity
value of 0.9, and the other surfaces emissivity value is 0.3, and the thermal conductivity
of the materials is summarized in Table 4.2 [4].
Thermal Finite Element Analysis | 49
Table 4.2 Thermal conductivities.
Material kx (W/°C/m) ky (W/°C/m) kz (W/°C/m)
Core laminated steel 31 31 0.6

Low voltage winding 7.04 7.04 7.04
High voltage winding 5.95 5.95 5.95
Figure 4.12 shows the temperature distribution output, it is possible to see that the
higher temperature spots are on the windings, and the surfaces closer to the windings. It
is interest to analyse the temperature gradient on z-direction due to the low thermal
conductivity of the laminated steel, if the thermal conductivity in z-direction was higher
the heat flow in this direction would increase, and consequently the temperature values
on plane xz would be smaller. Additionally, it is possible to observe that the temperature
distribution is symmetric on the planes illustrated in Figure 3.5, which supports the
boundary condition applied to the TRN model (section 3.4.3), where it was considered
that there is no heat transfer between the modelled portion of the transformer and its
symmetry counterpart, due to the minimal temperature difference and small heat path
between the modelled portion and its symmetry counterpart.
Figure 4.12 Transformer thermal FEA: temperature distribution.
The total heat flux presented in Figure 4.13, shows the paths of the heat exchange
occurring from the surfaces to the ambient and also the heat flowing from the windings
to the core, and from the core to the ambient. Thus, the core operates as a heat sink for
the heat generated in the windings.
Figure 4.13 Transformer thermal FEA: heat flux.
4.4 Coupled Electromagnetic and Thermal Finite Element

Analysis
The coupled analysis allows a complete interaction between the electromagnetic
and thermal fields. That is a natural demand to improve the design process of electric
machines to fulfil the nowadays requirements of power density and efficiency. The losses
obtained by the electromagnetic model are used as the heat generation input in the
thermal model, that will calculate the temperature distribution of the machine, and the
temperature output of the thermal model is used as input to the electromagnetic model,
in order to calculate the losses considering the temperature impact on the
electromagnetic parameters. This interaction between the electromagnetic and thermal
model runs iteratively until the variation of temperature between the two last iterations
is lower than a defined threshold, meaning that the temperature and losses variation is
minimal. Differently to the thermal model developed in section 4.3.1, where it is
considered a uniform heat generation, the coupled model considers the heat generation
distribution as the losses distribution output of the electromagnetic model.
The FEA regarding only one physics field was introduced in the previous sections.
So, the objective of this section is to couple the analyses that were previously treated
separately, therefore granting a complete interaction between the electromagnetic and
thermal fields. The use of coupled analysis is a natural demand to improve the design
process of electric machines to fulfil the nowadays requirements of power density and
efficiency.
The coupled analysis approach allows the evaluation of the electromagnetic model
according to the temperature change. Material properties from the magnetic and electric
circuits vary non linearly with temperature, and for this reason, the thermal model output
Coupled Electromagnetic and Thermal Finite Element Analysis | 51
can be crucial to evaluate the electromagnetic performance of the electric machines. The
temperature rise increases the resistive loss and also brings the risk of demagnetization
For instance, if the machine uses permanent magnets in its excitation system, for an
operating temperature of 100°C, the magnetic flux may be reduced till 20% when
compared with the magnetic flux at room temperature [2], which impacts directly on the
machine performance.
This approach is a handful in the machine design process, allowing the engineers
to evaluate the design choices impact in a wider view. Such as in [73], where the design
process of a medium frequency power transformer is done by utilizing FEA and the
validation was achieved by building the prototype based on the FEA project. Similarly,
[74] uses the FEA coupling to investigate different rotor structures of a synchronous
generator, observing that a claw pole rotor is more thermally efficient than a salient pole
rotor.
The coupled methodology proposed in this work is intended to be used as part of
the design process of an electric machine, or it can be useful to evaluate the machine
performance in operating conditions that differ from the rated ones. An example is the
derating of the machine in ambient temperature or altitude different from the ones
specified for the rated quantities may be more precise, if assisted with an electromagnetic
and thermal coupled approach.
Figure 4.14 summarizes the proposed coupled methodology. Most of the flowchart
elements were explained on the previous sections, thus this flowchart establishes the link
between the electromagnetic and thermal analysis. A magnetostatic solution is present in
this flowchart, since it is a simple and quick solution that allows the evaluation of some
features, such as iron saturation, hence guiding to some model changes e.g. geometries,
ferromagnetic material requirements, excitations sources. The generated losses resulted
from the electromagnetic FEA solution are used as heat generation in the thermal FEA.
This interaction between the solutions, grants the heat generation be closer as possible to
the real loss distribution. Due to the temperature change, material properties such as
electric conductivity will change, thus the electromagnetic solution needs to be updated
considering the components new temperatures. This iterative process runs until the
temperature difference between the current interaction and the previous minor than δ,
meaning that the temperatures and losses are stable.
Figure 4.14 Electromagnetic and thermal coupled analysis flowchart.
4.4.1 Numerical Coupled Modelling of a Low Power Transformer

By using Ansys Workbench 2019 R3 [75] an electromagnetic and thermal coupled
analysis of the single-phase transformer was made. The coupled analysis was set up by
importing the losses from Ansys Maxwell (electromagnetic FEA) into the Ansys
Mechanical (thermal FEA) and exporting the temperature distribution to the Ansys
Maxwell, and the thermal modifier was enabled in the electromagnetic FEA. This
interaction between the losses and the temperature was only applied to the resistive loss,
where the copper electric conductivity is defined as
Coupled Electromagnetic and Thermal Finite Element Analysis | 53
(4.1)
Where is the conductivity of the copper at 20 °C, that is equal to S/m, and α
is the copper thermal resistivity coefficient, which is equal to [76].
The electromagnetic and thermal FEA used are the models developed in section
4.2.1 and 4.3.1, respectively, with the difference that the internal heat generation is now
obtained from the Ansys Maxwell solution. Therefore, this simulation corresponds to the
transformer on its rated working conditions. In Figure 4.15 is possible to see the evolution
of the transformer losses along with the iterations. The resistive losses increased 32.82%
between the first iteration and the last, is possible to see that in the last three iterations
the losses increase were minimal.
Figure 4.15 Transformer coupled FEA: loss evolution.
The temperature rise is shown in Figure 4.16, which follows the same aspect of the
losses, in the initial iterations there is significant increase and in the last iterations, the
temperature rise is minimal. The average temperature of the model increased in 12.48°C
between the first and the last iteration, which is a significant temperature difference that
would be ignored if the coupled analysis was not applied.
Figure 4.16 Transformer coupled FEA: temperature evolution.

It is possible to see that for this case, was not needed to run several iterations, with
three iterations the simulation results were already stable. Were expected this low
number of iterations since only the resistive loss was considered temperature-dependent.
This analysis was done to simulate the transformer behaviour at electric rated
conditions, and under normal conditions of temperature and pressure. However, it is
possible to adjust the coupled FEA project to simulate different electric and ambient
conditions e.g. underload, overload, high ambient temperature, thin air. Thus, the coupled
analysis can assist the optimization and derating of the transformer design for
applications with such conditions.
4.5 Summary
In this chapter was discussed the numerical modelling of electric machines using the FEA,
which is a tool that was introduced decades ago in the structural analysis field, and
nowadays is developed for a variety of physics fields. FEA is a powerful tool that facilitates
the coupling of different physics fields, which can be crucial for the design process of
modern applications. Within the electric machine context, the electromagnetic and
thermal FEA and their coupling were addressed. And to demonstrate the methodology
presented for the numerical modelling, it was developed the numerical model for a single-
phase transformer.
Chapter 5
Results and Discussion
Experimental Results | 57
5 Results and Discussion

In the course of this work it was discussed different approaches to model electric
machines. These approaches were applied to a low power single-phase transformer. The
objective of this chapter is to summarize, discuss, compare, and validate the results
obtained from the models. First, the experimental procedure and measurements of the
transformer are presented, thereafter, the results from the models are summarized,
compared, and discussed.
5.1 Experimental Results

To obtain the experimental losses of the transformer, the device was submitted to two
tests, these tests were initiated with the transformer at ambient temperature (20 °C), as
these tests occur in a short period, the temperature rise of the machine can be ignored.
Appendix A contains more details about the performed tests. Therefore, the measured
resistive loss at 20 °C was 49.6 W, and the measured iron loss at 20 °C was 13 W.
The desired approach to measuring the temperature of the transformer was to use
the PT 100 temperature sensor in concerned spots of the transformer. Although, due to
the COVID-19 pandemic, it was not possible to perform this procedure. In order to still
have experimental validation, a student under Prof. Ângela Ferreira supervision during
his internship period in IPB, which occurred previous to this work, obtained the
experimental thermal data used in this work.
The temperature monitoring of the transformer was accomplished by using an
infrared thermography camera. The camera absorbs the infrared radiation emitted by the
body, and by using the Stefan-Boltzman’s law the body temperature is obtained [70]. Its
use allows measuring the surfaces temperatures of the device without contact. Thus, it is
not possible to measure the temperature of spots that are not in the device surface, such
as the temperature between the windings.
In the laboratory at the ambient temperature of 20 °C, the transformer was
submitted to the rated working condition. Thus, the high voltage winding was excited by
the rated voltage, and on the low voltage winding, a resistive load was connected..
Thermal behaviour of the machine was supervised through an infrared camera. Figure 5.1
shows the thermal steady-state of the transformer. The infrared camera was configured
with the emissivity of the black insulation tape of the windings. The obtained images are
58 | Results and Discussion
useful to evaluate the temperature gradient and the thermal steady state condition, not
being able to evaluate absolute temperatures of the surfaces.
Figure 5.1 Thermal image of the transformer after 200 minutes.
As the thermal images are not reliable to obtain absolute temperatures, the
temperature of the windings was measured by using the procedure recommended by the
IEC 60034 [77]. Which consists in measuring the winding resistance at the known initial
temperature and measuring the winding resistance after the temperature rise, after these
measurements, the temperature of the winding is determined by
(5.1)
where Rf is the winding resistance at the final temperature, Ri the winding resistance at
the initial temperature Ti, and λ is the reciprocal of the temperature coefficient of
resistance at 0 °C of the conductor material, that for copper is 235 and 225 for aluminium.
Table 5.1 shows the results obtained from the IEC method, with the values of the
resistances of the windings at steady-state temperature, the resistive loss can be
calculated considering the temperature rise in the windings, the skin-effect is ignored in
this case since the skin depth of the copper is higher than the diameter of the winding’s
wire.
Table 5.1 Windings temperature measurement (IEC 60034).
Ri (Ω) Rf (Ω) Tf (°C) Resistive Loss @Tf (W)

Low voltage winding 0.4 0.52 96.5 36.26
High voltage winding 1.0 1.3 96.5 25.20
Summary of the Models Results and Discussion | 59
5.2 Summary of the Models Results and Discussion

5.2.1 Losses Results
In section 4.2.1 the electromagnetic FEA model of the transformer was developed with
the objective to calculate the losses of the machine. With losses measurements of the
transformer, it is possible to evaluate the accuracy of the loss prediction of the model.
Table 5.2 shows the losses measured and calculated, and the relative error between them.
These results are related to losses of the transformer at ambient temperature of 20 °C.
Table 5.2 Transformer losses: experimental and FEA.
Resistive Loss (W) Iron Loss (W)

FEA 44.12 13.78
Experimental 49.60 13.00
Error (%) 11.05 6.00
The results obtained from the model are in agreement with the experimental
measures. The errors presented by the model may be related to some uncertainty in data
of the active materials and dimensions of the physical transformer, e.g., number of turns
of each winding, specifications of the ferromagnetic material used on the core. As
explained in section 2.1.2 the prediction of the iron loss is very difficult due to the
complexity of the phenomena involved. The advantage of using an FEA model is that the
iron loss is calculated for each element, considering the non-uniform flux density on the
core. Given these aspects, the results presented by the model are acceptable.
5.2.2 Temperature Results

The analytical and numerical thermal model of the transformer was developed in section
3.4 and 4.3.1, respectively. Both methods use the losses obtained by the experimental test
at 20 °C, considering a uniform distribution of the iron loss on the core, and the resistive
loss on the windings. Figure 5.2 shows the temperature spots used to compare the
temperature results of the analytical and numerical model.
Figure 5.2 Transformer temperature spots
Table 5.3 shows the results from both models and their difference. The
temperature spots T2 and T3 presented a small difference between the models, whereas
the spots T1 and T4 presented a temperature difference of about 10%. The discretization
levels between the models are very different, which can be related to the temperature
difference between the models. The TRN approach used a total of 26 volume elements to
model ¼ of the transformer, whereas the FEA used 15821 elements to model it entirely.
Even with this discrepancy between the discretization level, the results are similar,
although this may not be the case for machines with complex geometries. Another factor
that impacts the results difference is related to the convection coefficients. The model
used in TRN considered the convection coefficient constant. In contrast, the numerical
approach considered its variation with the surface temperature, which leads to a more
precise convection heat transfer calculation in the FEA method.
Table 5.3 Temperature results from the transformer thermal models
Spot TRN (°C) FEA (°C) Difference (%)

T1 102 91.52 11.45
T2 81.8 81.92 -0.14
T3 91.5 88.01 3.97
T4 70.9 78.04 -9.14
In section 4.4.1, it was developed the numerical coupled electromagnetic and

thermal model of the transformer, where the models run iteratively, considering the
temperature dependency on the loss calculation and the non-uniform heat generation.
When the solution reaches the established threshold of the temperature difference
between the two last iterations, the temperature output of the coupled model
corresponds to the transformer at its steady-state condition. Therefore, the windings’
temperature output of the model on the last iteration can be compared to the
Summary of the Models Results and Discussion | 61
temperatures measured through the IEC 60034 method after the transformer reached a
steady-state condition.
The results of the analytical and numerical thermal model previously shown, are
obtained considering the losses of the transformer at ambient temperature. Thus, they
can not be compared to the temperature measured on steady-state conditions. So, the
resistive loss of the models was calculated and updated considering the windings’
resistance at steady-state temperature. Table 5.1 shows the resistive loss at the steady-
state temperature.
Table 5.4 shows the steady-state temperature of the windings, that was obtained
experimentally through the IEC 60034 method, and the temperature of the models
corresponds to the average temperature of the windings. The TRN model presented an
error of about 20% in relation to the experimental value, whereas the FEA and coupled
FEA models presented errors of around 7%.
Table 5.4 Temperature of the windings at steady-state
Experimental TRN FEA Coupled

TRN FEA Coupled
Location (IEC 60034) Error Error FEA Error
(°C) (°C) FEA (°C)
(°C) (%) (%) (%)
Low
voltage 96.5 115.6 19.51 102.89 6.62 103.57 7.3
winding
High
voltage 96.5 116.17 20.38 101.95 5.65 102.91 6.64
winding
Table 5.5 shows the temperature results of the models at the steady-state
condition regarding the temperature spots indicated by Figure 5.2, and the temperature
difference between the models. Reiterating, that there is no experimental value for these
temperature spots, because the infrared camera does not provide reliable measurements
of the absolute temperature on the spots.
Table 5.5 Transformer temperature results: TRN, FEA, and coupled FEA
Temperature Difference (%)

TRN FEA Coupled FEA
Location TRN vs TRN vs FEA vs
(°C) (°C) (°C)
FEA Coupled FEA Coupled FEA
T1 115.00 101.32 101.27 13.50 13.56 0.05
T2 89.20 88.27 88.13 1.05 1.21 0.16
T3 100.00 96.62 100.41 3.50 -0.41 -3.77
T4 76.70 84.93 85.01 -9.69 -9.79 -0.09
The results show that the FEA and coupled FEA models presented results very
similar results, the highest temperature difference between them occurs on spot T3 where
the FEA model resulted on a temperature 3.77% cooler than the coupled FEA. This small
difference between the models is related to the losses used on the models. The thermal
FEA uses the losses obtained experimentally, and the use of resistive loss at steady-state
temperature was only possible due to the measurement of the windings’ resistance at the
steady-state temperature. While, the coupled model uses the losses obtained by the
electromagnetic model, which are calculated automatically considering the non-uniform
loss distribution and the temperature dependency of the materials. Thus, the coupled FEA
model does not require the physical prototype to obtain the losses of the machine for
different temperatures. And this is why the coupled model facilitates the design process
of electric machines, without the need of a physical prototype is possible to simulate the
multiphysics behaviour of the device, allowing to study design choices without the time
and cost of building numerous prototypes.
The results from the TRN are very similar to the numerical methods on the
temperature spots T2 and T3, while presented significant difference on spots T1 and T4.
These results indicate that the higher level of discretization present on the numerical
methods allowed better modelling of the heat paths on the windings. Indicating that the
thermal resistances of the windings were overestimated. Besides that, the convection heat
transfer is calculated more precisely on the numerical methods, considering the
temperature of the surface on the value of the convection heat transfer coefficient.
With the results presented by Table 5.4 and Table 5.5, it is possible to draw some
conclusions regarding the performance and improvements of the models. All models
predicted a higher temperature on the windings than the ones obtained experimentally.
Thus, it should be considered the use of experimental thermal coefficients of the actual
Summary | 63
materials and fluid flow of the transformer, which would allow the models to better
resemble the physical transformer.
5.3 Summary
The modelling approaches of electric machines discussed during this work were applied
to a low power single-phase transformer, due to its simplicity and availability on the
laboratory. This chapter summarized the results obtained from the developed models and
confronted the theoretical results with experimental results obtained on the laboratory.
The numerical models presented good performance, but the analytical thermal model
presented significant error when compared to the experimental measurements. With the
results presented allows identifying future improvements for the models, such as the
characterization of the electromagnetic and thermal parameters of the actual materials
present on the transformer. Also, the TRN results suggest an overestimation of the
thermal resistance, especially on the windings, and should be considered a more refined
calculation of the convection heat transfer.
Chapter 6
Conclusion
Summary | 67
6 Conclusion
This work aimed at discussing the modelling process of electric machines focusing on the
electromagnetic and thermal modelling. The classic design process of electric machines is
focused on the electromagnetic design and overlooks the thermal design, relying on the
use of empirical experience, especially for medium and small-sized machines. As a
reaction to the modern requirements of electric machines applications, the thermal
modelling of the electric machine has been receiving more attention in the past years.
The thermal model is achieved by an analytical and numerical approach. The
analytical approach, TRN, discretizes the heat transfer phenomena into thermal
resistances, similar to an electric circuit. Regarding the numerical model, it is obtained by
FEA that is commonly used due to its high level of discretization granting a precise
temperature distribution. Both methods were applied to model a low power transformer,
where the FEA model presented better results on the prediction of the temperature of the
windings when compared to experimental measurements. Results showed that the higher
level of discretization present in the FEA model has a significant impact on the
determination of the hot spots. Also, the FEA model has a straightforward implementation
of the temperature dependency of the convection heat transfer coefficient, allowing to
model more accurately the heat transfer through convection. The TRN model shows that
despite being a simple analytical method, it can produce satisfactory results.
Electric machines temperature rise is caused by the device’s losses, which are
mainly electromagnetic losses, being divided into resistive loss and iron loss. The resistive
loss is usually the principal heat source in electric machines, the electric resistance of the
conductor is strongly dependent on its temperature, this relation is expressed by a simple
equation, allowing a direct temperature dependency of the resistive loss. Whereas the
iron loss modelling is a difficult challenge itself, having different approaches and
formulations, therefore its thermal dependency it is not straightforward, though, is
known that the iron loss reduces with the temperature rise. Also, the fundamentals of heat
transfer were established, which allowed the understanding of the main aspects related
to heat removal mechanisms.
A methodology for coupling the electromagnetic and thermal analyses was
proposed to handle the multiphysics characteristics of electric machines. By applying the
coupled analysis approach to the low power transformer, it is possible to observe the
68 | Conclusion
temperature rise due to the losses generation and the increase of the resistive loss due to
the temperature rise in the windings. Besides that, the losses obtained by the
electromagnetic model corresponds to a non-uniform heat generation accordingly to the
magnetic flux density in the core and current density distribution in the windings This
coupled methodology was developed to assist in the design process and optimization of
electric machines, and also in the derating process for conditions different from the rated
ones, for instance, higher ambient temperature, different levels of altitude and other
factors that may affect the device performance and lifetime. Therefore, with this analysis
methodology, it is possible to obtain a wider view of the machine performance without
the time and financial resources of building various prototypes.
By using a case study of a simple electric machine, it was possible to understand
the methods’ main advantages and disadvantages. Moreover, from the results obtained
the models presented satisfactory results when compared to the experimental data.
This work dealt with the thermal model of electric machines that has been
receiving more attention in the past years, also presented a coupled analysis
methodology, that is a trending method due to the advances in the computer-aided
engineer software.
6.1 Future Work

The first issues to improve is the experimental thermal data of the transformer. As
discussed in chapter 5, the thermal images do not give useful information about the
transformer’s core temperature and only provides the surfaces’ temperature. Therefore,
the objective is to use of thermal sensors, for instance, PT100, to measure the temperature
of relevant spots, e.g., between windings and, between windings and core. Also, there is
room for improvement in the thermal models, especially in the TRN model, which
presented significant error when compared to the experimental temperature of the
windings. The use of experimental thermal coefficients rather than empirical ones should
increase the accuracy of the model, besides a more refined calculation of the convection
on the TRN model
As the electromagnetic and thermal coupling model simulates a more realistic
operation of the device, the use of the coupled model to build a digital twin would enable
to exploit other important aspects of the device beyond the design process. A digital twin
can be defined as computer-based models that are simulating, emulating, mirroring the
Future Work | 69
life of a physical entity, a digital twin is the virtual counterpart of a physical object [78].
That in the context of this work would be the electric machine modelled. Helping in the
monitoring of the device in its application, enabling to keep track on the ageing factor of
the materials, such as the insulation materials, also allowing to manage the predictive
maintenance better and predict faults [79]–[81], which can be crucial for applications
where the maintenance programs are of difficult implementation [82], [83].
Furthermore, future work should apply the electromagnetic and thermal coupled
analysis to other electric machines, e.g., permanent magnet synchronous machines,
induction motors, high power transformers. In addition, a more significant study on the
iron loss modelling should consider its temperature dependency.
Bibliography | 71
Bibliography
[1] IEA, “World electricity final consumption by sector, 1974-2018,” 2020.
https://www.iea.org/data-and-statistics/charts/world-electricity-final-
consumption-by-sector-1974-2018 (accessed Oct. 27, 2020).
[2] R. Marius et al., Multiphysics Simulation by Design for Electrical Machines, Power
Electronics, and Drives. John Wiley & Sons, 2018.
[3] E. Schmidt, “Finite element analysis of electrical machines and transformers: State
of the art and future trends,” COMPEL - Int. J. Comput. Math. Electr. Electron. Eng.,
vol. 30, no. 6, pp. 1899–1913, 2011, doi: 10.1108/03321641111168183.
[4] J. Pyrhönen, T. Jokinen, and V. Hrabovcová, Design of Rotating Electrical Machines.
John Wiley & Sons, 2008.
[5] D. Staton, A. Boglietti, and A. Cavagnino, “Solving the more difficult aspects of
electric motor thermal analysis,” IEMDC 2003 - IEEE Int. Electr. Mach. Drives Conf.,
vol. 2, no. 3, pp. 747–755, 2003, doi: 10.1109/IEMDC.2003.1210320.
[6] M. Popescu et al., “Modern Heat Extraction Systems for Electrical Machines – A
Review,” pp. 289–296, 2015.
[7] G. Du, W. Xu, N. Huang, X. Cheng, and X. Xiao, “Rotor Design of High Power High
Speed Permanent Magnet Machine Considering Multiphysics Constraints,” 2019
22nd Int. Conf. Electr. Mach. Syst. ICEMS 2019, vol. 1, 2019, doi:
10.1109/ICEMS.2019.8921870.
[8] Z. Huang and J. Fang, “Multiphysics Design and Optimization of High-Speed
Permanent-Magnet Electrical Machines for Air Blower Applications,” IEEE Trans.
Ind. Electron., vol. 63, no. 5, pp. 2766–2774, 2016, doi: 10.1109/TIE.2016.2518121.
[9] Y. Yasa, M. Elamin, Y. Sozer, J. Kutz, J. S. Tylenda, and R. L. Wright, “Acoustic noise
mitigation for high pole count switched reluctance machines through skewing
method with multiphysics FEA simulations,” 2017 IEEE Energy Convers. Congr.
Expo. ECCE 2017, vol. 2017-Janua, no. 1, pp. 738–744, 2017, doi:
10.1109/ECCE.2017.8095858.
[10] C. Yan, Y. Wei, C. Xu, and B. Zhang, “Experimental verification and electromagnetic-
mechanics-acoustic field coupling analysis of transformer pressure relief valve
malfunctions due to external short-circuit faults,” IEEE Trans. Magn., vol. 55, no. 6,
pp. 9–12, 2019, doi: 10.1109/TMAG.2019.2900660.
72 | Bibliography
[11] T. Ohinata and K. Arimatsu, “Reluctance Network Model of Three-Phase-

Laminated-Core,” IEEE Trans. Magn., vol. 55, no. 7, pp. 1–6, 2019, doi:
10.1109/TMAG.2019.2900527.
[12] E. M. Barhoumi, F. Wurtz, C. Chillet, and B. Ben Salah, “Reluctance network model
for linear switched reluctance motor,” 12th Int. Multi-Conference Syst. Signals
Devices, SSD 2015, pp. 1–4, 2015, doi: 10.1109/SSD.2015.7348156.
[13] A. A. Diriye, Y. Amara, and G. Barakat, “Three-Dimensional Modeling of Permanents
Magnets Synchronous Machines Using a 3D Reluctance Network,” Proc. - 2018 23rd
Int. Conf. Electr. Mach. ICEM 2018, pp. 2304–2310, 2018, doi:
10.1109/ICELMACH.2018.8506820.
[14] C. Freitag, “Magnetic Properties of Electrical Steel, Power Transformer Core Losses
and Core Design Concepts,” Elektrotechnik und Informationstechnik des
Karlsruher Institutes für Technologie (KIT), 2017.
[15] L. Susnjic, Z. Haznadar, and Z. Valkovic, “3D ﬁnite-element determination of stray
losses in power transformer,” Electr. Power Syst. Res. J., vol. 78, pp. 1841–1818,
2008.
[16] S. Xue, J. Feng, S. Guo, J. Peng, W. Q. Chu, and Z. Q. Zhu, “A new iron loss model for
temperature dependencies of hysteresis and eddy current losses in electrical
machines,” IEEE Trans. Magn., vol. 54, no. 1, 2018, doi:
10.1109/TMAG.2017.2755593.
[17] D. Lin, P. Zhou, W. N. Fu, Z. Badics, and Z. J. Cendes, “A Dynamic Core Loss Model for
Soft Ferromagnetic and Power Ferrite Materials in Transient Finite Element
Analysis,” 1318 IEEE Trans. Magn., vol. 40, no. 2, pp. 1318–1321, 2004.
[18] A. Boglietti, A. Cavagnino, D. Staton, M. Shanel, M. Mueller, and C. Mejuto, “Evolution
and modern approaches for thermal analysis of electrical machines,” IEEE Trans.
Ind. Electron., vol. 56, no. 3, pp. 871–882, 2009, doi: 10.1109/TIE.2008.2011622.
[19] T. A. Jankowski et al., “Development and validation of a thermal model for electric
induction motors,” IEEE Trans. Ind. Electron., vol. 57, no. 12, pp. 4043–4054, 2010,
doi: 10.1109/TIE.2010.2043044.
[20] A. Boglietti, A. Cavagnino, M. Lazzari, and M. Pastorelli, “A Simplified Thermal Model
for Variable-Speed Self-Cooled Industrial Induction Motor,” IEEE Trans. Ind. Appl.,
vol. 39, no. 4, pp. 945–952, 2003, doi: 10.1109/TIA.2003.814555.
[21] M. Cavazzuti, G. Gaspari, S. Pasquale, and E. Stalio, “Thermal management of a
Bibliography | 73
Formula E electric motor: Analysis and optimization,” Appl. Therm. Eng., vol. 157,
no. May, p. 113733, 2019, doi: 10.1016/j.applthermaleng.2019.113733.
[22] S. Nategh, H. Zhang, O. Wallmark, A. Boglietti, T. Nassen, and M. Bazant, “Transient
Thermal Modeling and Analysis of Railway Traction Motors,” IEEE Trans. Ind.
Electron., vol. 66, no. 1, pp. 79–89, 2019, doi: 10.1109/TIE.2018.2821619.
[23] Y. Xie and Y. Wang, “3D temperature field analysis of the induction motors with
broken bar fault,” Appl. Therm. Eng., vol. 66, no. 1–2, pp. 25–34, 2014, doi:
10.1016/j.applthermaleng.2014.02.008.
[24] M. Allahbakhshi and M. Akbari, “Heat analysis of the power transformer bushings
using the finite element method,” Appl. Therm. Eng., vol. 100, pp. 714–720, 2016,
doi: 10.1016/j.applthermaleng.2016.02.065.
[25] D. Sarkar and A. K. Naskar, “Computation of thermal condition in an induction
motor during reactor starting,” Int. J. Electr. Power Energy Syst., vol. 44, no. 1, pp.
938–948, 2013, doi: 10.1016/j.ijepes.2012.08.004.
[26] E. Galloni, P. Parisi, F. Marignetti, and G. Volpe, “CFD analyses of a radial fan for
electric motor cooling,” Therm. Sci. Eng. Prog., vol. 8, no. September, pp. 470–476,
2018, doi: 10.1016/j.tsep.2018.10.003.
[27] M. Satrústegui, G. Artetxe, I. Elosegui, M. Martinez-Iturralde, and J. C. Ramos,
“Wafter design for totally enclosed electric machines,” Appl. Therm. Eng., vol. 129,
pp. 93–105, 2018, doi: 10.1016/j.applthermaleng.2017.10.011.
[28] J. Pyrhönen, P. Lindh, M. Polikarpova, E. Kurvinen, and V. Naumanen, “Heat-transfer
improvements in an axial-flux permanent-magnet synchronous machine,” Appl.
Therm. Eng., vol. 76, pp. 245–251, 2015, doi:
10.1016/j.applthermaleng.2014.11.003.
[29] A. . Fitzgerald, Electric Machinery, Sixth Edition, 6th ed. McGraw-Hill, 2003.
[30] M. N. O. Sadiku, Elements of Electromagnetics, Sixth. Oxford University Press, 2015.
[31] C. D. Graham, “Physical origin of losses in conducting ferromagnetic materials,” J.
Appl. Phys., vol. 53, no. August, 1982, doi: 10.1063/1.330306.
[32] S. N. Vukosavic, Electrical Machines, 1st ed. Springer, 2013.
[33] D. Eggers, S. Steentjes, and K. Hameyer, “Advanced iron-loss estimation for
nonlinear material behavior,” IEEE Trans. Magn., vol. 48, no. 11, pp. 3021–3024,
2012, doi: 10.1109/TMAG.2012.2208944.
[34] W. Roshen, “Iron loss model for permanent-magnet synchronous motors,” IEEE
74 | Bibliography
Trans. Magn., vol. 43, no. 8, pp. 3428–3434, 2007, doi:

10.1109/TMAG.2007.899687.
[35] Giorgio Bertotti, “General properties of power losses in soft ferromagnetic
materials,” IEEE Trans. Magn., vol. 24, no. 1, pp. 621–630, 1988.
[36] A. Boglietti, A. Cavagnino, M. Lazzari, and M. Pastorelli, “Predicting iron losses in
soft magnetic materials with arbitrary voltage supply: An engineering approach,”
IEEE Trans. Magn., vol. 39, no. 2 II, pp. 981–989, 2003, doi:
10.1109/TMAG.2003.808599.
[37] W. A. Roshen, “A practical, accurate and very general core loss model for
nonsinusoidal waveforms,” IEEE Trans. Power Electron., vol. 22, no. 1, pp. 30–40,
2007, doi: 10.1109/TPEL.2006.886608.
[38] J. Reinert, A. Brockmeyer, and R. W. A. A. De Doncker, “Calculation of losses in ferro-
and ferrimagnetic materials based on the modified Steinmetz equation,” IEEE
Trans. Ind. Appl., vol. 37, no. 4, pp. 1055–1061, 2001, doi: 10.1109/28.936396.
[39] M. Albach, T. Duerbaum, and A. Brockmeyer, “Calculating core losses in
transformers for arbitrary magnetizing currents a comparison of different
approaches,” PESC Rec. - IEEE Annu. Power Electron. Spec. Conf., vol. 2, pp. 1463–
1468, 1996, doi: 10.1109/pesc.1996.548774.
[40] D. C. Hanselman, Brushless Permanent Magnet Motor Design, 2nd ed., vol. 26, no. 3.
The Writer’s Collective, 2006.
[41] K. Yamazaki and Y. Kato, “Iron loss analysis of interior permanent magnet
synchronous motors by considering mechanical stress and deformation of stators
and rotors,” IEEE Trans. Magn., vol. 50, no. 2, pp. 22–25, 2014, doi:
10.1109/TMAG.2013.2282187.
[42] M. Nakano et al., “High-Precision calculation of iron loss by considering stress
distribution of magnetic core,” IEEJ Trans. Ind. Appl., vol. 129, no. 11, pp. 1060–
1067, 2009, doi: 10.1541/ieejias.129.1060.
[43] M. Satoh, S. Kaneko, M. Tomita, S. Doki, and S. Okuma, “Stator shape improvement
to decrease iron loss caused by shrink fit utilizing physical properties of electrical
steel sheet,” IEEJ Trans. Ind. Appl., vol. 127, no. 1, pp. 60–68, 2007, doi:
10.1541/ieejias.127.60.
[44] J. Chen, D. Wang, Y. Jiang, X. Teng, S. Cheng, and J. Hu, “Examination of Temperature-
Dependent Iron Loss Models Using a Stator Core,” IEEE Trans. Magn., vol. 54, no. 11,
Bibliography | 75
2018, doi: 10.1109/TMAG.2018.2829632.

[45] J. Driesen, R. Belmans, and K. Hameyer, “Methodologies for coupled transient
electromagnetic-thermal finite element modeling of electrical energy transducers,”
IEMDC 2001 - IEEE Int. Electr. Mach. Drives Conf., pp. 681–686, 2001, doi:
10.1109/IEMDC.2001.939387.
[46] F. P. Incropera, D. P. Dewitt, T. L. Bergman, and A. S. Lavine, Fundamentals of Heat
and Mass Transfer, 6th ed. John Wiley & Sons, 2006.
[47] Y. A. Cengel, Heat Transfer A Practical Approach, 2nd ed. Mcgraw-Hill, 2002.
[48] M. Popescu, D. A. Staton, A. Boglietti, A. Cavagnino, D. Hawkins, and J. Goss, “Modern
Heat Extraction Systems for Power Traction Machines - A Review,” IEEE Trans. Ind.
Appl., vol. 52, no. 3, pp. 2167–2175, 2016, doi: 10.1109/TIA.2016.2518132.
[49] M. Li, Y. An, and Z. Zhang, “Improving Thermal Analysis Accuracy of LPTN for
Vehicle Claw-Pole Alternator by Calibrating Forced Convection Coefficients Based
on Experimental Results,” IEEE Access, vol. 7, pp. 129327–129334, 2019, doi:
10.1109/ACCESS.2019.2940508.
[50] R. Shafaie, M. Kalantar, and A. Gholami, “Thermal Analysis of 10-MW-Class Wind
Turbine HTS Synchronous Generator,” vol. 24, no. 2, 2014.
[51] M. Jaritz and J. Biela, “Analytical model for the thermal resistance of windings
consisting of solid or litz wire,” 2013 15th Eur. Conf. Power Electron. Appl., pp. 1–10,
2013, doi: 10.1109/EPE.2013.6634624.
[52] R. Wrobel, S. Ayat, and J. L.Baker, “Analytical Methods for Estimating Equivalent
Thermal Conductivity in Impregnated Electrical Windings.pdf,” 2017 IEEE Int.
Electr. Mach. Drives Conf., pp. 1–8, 2017, doi: 10.1109/IEMDC.2017.8002003.
[53] H. Liu, S. Ayat, R. Wrobel, and C. Zhang, “Comparative study of thermal properties
of electrical windings impregnated with alternative varnish materials,” J. Eng., vol.
2019, no. 17, pp. 3736–3741, 2019, doi: 10.1049/joe.2018.8198.
[54] A. Boglietti et al., “Equivalent Thermal Conductivity Determination of Winding
Insulation System by Fast Experimental Approach,” 2015 IEEE Int. Electr. Mach.
Drives Conf., pp. 1215–1220, 2015, doi: 10.1109/IEMDC.2015.7409216.
[55] S. Ayat, R. Wrobel, J. Goss, and D. Drury, “Equivalent Thermal Conductivity for
Impregnated Electrical Windings Formed from Profiled Rectangular Conductors .
In 8th IET International Conference on Power Electronics , Machines and Drives (
PEMD 2016 ) Estimation of Equivalent Thermal Conductivity fo,” 8th IET Int. Conf.
76 | Bibliography
Power Electron. Mach. Drives (PEMD 2016), pp. 1–6, 2016, doi:
10.1049/cp.2016.0313.
[56] J. . Holman, Heat Transfer, 10th ed. Mcgraw-Hill, 2010.
[57] A. Boglietti, A. Cavagnino, M. Parvis, S. Member, and A. Vallan, “Evaluation of
Radiation Thermal Resistances in Industrial Motors,” IEEE Trans. Ind. Appl., vol. 42,
no. 3, pp. 688–693, 2006, doi: 10.1109/TIA.2006.873655.
[58] NI, “Multisim.” 2015.
[59] P. Dechaumphai and S. Sucharitpwatskul, Finite Element Analysis with ANSYS
Workbench. Alpha Science International LTD, 2018.
[60] M. N. O. Sadiku, Numerical Techniques in Electromagnetics, 2nd ed. CRC Press, 2000.
[61] İ. H. Teke, Y. Özüpak, M. S. Mamiş, E. E. B. Thermal, and P. Plant, “Electromagnetic
Field and Total Loss Analysis of Transformers by Finite Element Method Abstract :,”
vol. 8, no. 1, pp. 24451–24460, 2019, doi: 10.18535/ijecs/v8i1.01.
[62] M. Yazdani-Asrami, M. Mirzaie, and A. A. Shayegani Akmal, “No-load loss calculation
of distribution transformers supplied by nonsinusoidal voltage using three-
dimensional finite element analysis,” Energy, vol. 50, no. 1, pp. 205–219, 2013, doi:
10.1016/j.energy.2012.09.050.
[63] D. M. Ionel, M. Popescu, S. J. Dellinger, T. J. E. Miller, R. J. Heideman, and M. I. McGilp,
“On the variation with flux and frequency of the core loss coefficients in electrical
machines,” IEEE Trans. Ind. Appl., vol. 42, no. 3, pp. 658–667, 2006, doi:
10.1109/TIA.2006.872941.
[64] S. Kül, O. Bilgin, and M. Mutluer, “Application of Finite Element Method to
Determine the Performances of the Line Start Permanent Magnet Synchronous
Motor,” Procedia - Soc. Behav. Sci., vol. 195, pp. 2586–2591, 2015, doi:
10.1016/j.sbspro.2015.06.458.
[65] L. Hu, K. Yang, C. Suo, Y. Ding, and W. Yu, “Analysis of Radial Electromagnetic Force
in Permanent Magnet Machine with Manufacturing Tolerance,” ICEMS 2018 - 2018
21st Int. Conf. Electr. Mach. Syst., pp. 256–260, 2018, doi:
10.23919/ICEMS.2018.8549530.
[66] H. H. Hanafy, T. M. Abdo, and A. A. Adly, “2D finite element analysis and force
calculations for induction motors with broken bars,” Ain Shams Eng. J., vol. 5, no. 2,
pp. 421–431, 2014, doi: 10.1016/j.asej.2013.11.003.
[67] Ansys, “Ansys® Academic Research Maxwell, Release 2019 R3.” 2019.
Bibliography | 77
[68] UGINE ACIERS DE CHATILLON ET GUEUGNON, “Fe V 470 - 50 HA.” 2011.

[69] ANSYS, “Maxwell Help,” 2019. www.ansys.com (accessed Mar. 26, 2020).
[70] T. Matsushita, Electricity and magnetism : new formulation by introduction of
superconductivity. 2014.
[71] R. Sitar, I. Šulc, and Ž. Janić, “Prediction of local temperature rise in power
transformer tank by FEM,” Procedia Eng., vol. 202, pp. 231–239, 2017, doi:
10.1016/j.proeng.2017.09.710.
[72] Ansys, “Ansys® Academic Research Mechanical, Release 2019 R3.” 2019.
[73] S. Balci, I. Sefa, and N. Altin, “Design and analysis of a 35 kVA medium frequency
power transformer with the nanocrystalline core material,” Int. J. Hydrogen Energy,
vol. 42, no. 28, pp. 17895–17909, 2017, doi: 10.1016/j.ijhydene.2017.03.158.
[74] D. Arumugam, P. Logamani, and S. Karuppiah, “Electromagnetic & thermal analysis
of synchronous generator with different rotor structures for aircraft application,”
Alexandria Eng. J., vol. 57, no. 3, pp. 1447–1457, 2018, doi:
10.1016/j.aej.2017.03.020.
[75] I. ANSYS, “Ansys® Academic Research Workbench, Release 2019 R3.” 2019.
[76] G. LWW, “Technical data for Winding Wire.” http://www.lww.se/wp-
content/uploads/2017/06/LWW-brochure-2016.pdf (accessed Feb. 10, 2020).
[77] IEC, IEC 60034-1 Rotating Electrical Machines - Rating and Performance. IEC, 2017.
[78] B. R. Barricelli, E. Casiraghi, and D. Fogli, “A survey on digital twin: Definitions,
characteristics, applications, and design implications,” IEEE Access, vol. 7, no. Ml, pp.
167653–167671, 2019, doi: 10.1109/ACCESS.2019.2953499.
[79] J. Wang, L. Ye, R. X. Gao, C. Li, and L. Zhang, “Digital Twin for rotating machinery
fault diagnosis in smart manufacturing,” Int. J. Prod. Res., vol. 57, no. 12, pp. 3920–
3934, 2019, doi: 10.1080/00207543.2018.1552032.
[80] P. Aivaliotis, K. Georgoulias, and G. Chryssolouris, “The use of Digital Twin for
predictive maintenance in manufacturing,” Int. J. Comput. Integr. Manuf., vol. 32, no.
11, pp. 1067–1080, 2019, doi: 10.1080/0951192X.2019.1686173.
[81] Y. Yang et al., “State Evaluation of Power Transformer Based on Digital Twin,” Proc.
- IEEE Int. Conf. Serv. Oper. Logist. Informatics 2019, SOLI 2019, pp. 230–235, 2019,
doi: 10.1109/SOLI48380.2019.8955043.
[82] K. Sivalingam, M. Sepulveda, M. Spring, and P. Davies, “A Review and Methodology
Development for Remaining Useful Life Prediction of Offshore Fixed and Floating
78 | Bibliography
Wind turbine Power Converter with Digital Twin Technology Perspective,” Proc. -
2018 2nd Int. Conf. Green Energy Appl. ICGEA 2018, pp. 197–204, 2018, doi:
10.1109/ICGEA.2018.8356292.
[83] S. S. Johansen, “On Developing a Digital Twin for Fault Detection in Drivetrains of
Offshore Wind Turbines,” no. June, 2018.
Appendix A
Transformer Characterization
Transformer Characterization | 81
A. Transformer Characterization
In this appendix the low power shell-type single-phase transformer used during this work
as case study will be detailed. This transformer was chosen as the object of study due to
its simplicity, not having moving parts and its availability on the laboratory. Figure A.1
shows the device used and its nameplate.
Figure A.1 Single-phase transformer (a) and its nameplate (b).
Figure A.2 shows the configuration of the windings, that was determined by
analysing the windings’ wire diameter since the rated current of the low voltage winding
is higher than the rated current of the high voltage winding, the wire with a larger
diameter will correspond to the low voltage winding, which was present on the outer
winding, thus the high voltage winding is inner winding. It was possible to measure only
the diameter of the low voltage winding that is 2 mm. In the device’s nameplate was
identified the rated electric specifications, the rated power is 1 kW, and the rated voltage
of 220 V for the primary winding and 110 V to the secondary winding, hence the rated
transformation ratio is 2:1.
Figure A.2 Windings configuration.

82 | Transformer Characterization
To obtain the device’s resistive and iron losses at room temperature (20 °C) a
short-circuit and a no-load test were performed. Table A.1 summarizes the experimental
data measured. The short-circuit test was accomplished by supplying the primary
winding with its rated current and the short-circuiting the secondary winding, with this
configuration the voltage applied in the primary winding is minimal, therefore the iron
loss can be ignored and the power measured corresponds to the resistive loss at nominal
load. In the no-load test, the primary winding was excited by its rated voltage (220 V) and
the secondary winding was in open-circuit configuration, since the current flowing
through the winding is small the resistive loss is minimal, therefore the power measured
in this test corresponds to the transformer’s iron loss. Also, in the no-load test, it is
measured the real transformation ratio that is 220:116.
Table A.1 Transformer tests data at room temperature.
V1 (V) I1 (A) V2 (V) I2(A) P (W) R1 (Ω) R2 (Ω)

Short-circuit 9.62 4.51 - 8.55 49.60
1.0 0.4
No-load 220 174 × 10−3 116 - 13
Figure A.3 Low voltage winding cross-section.
The number of turns of each winding was not specified by the manufacturer,
therefore it is necessary to estimate it. It is known the wire’s diameter of the low voltage
winding, assuming that the number of turns fills the low voltage winding cross-sectional
area, as shown in Figure A.3, the number of turns for the low voltage winding is given by,
(A.1)
where r2 is the wire radius of the winding, which results in , and is
obtained by using the real transformation ratio.
Transformer Characterization | 83
Figure A.4 shows the geometry parameters used to model the transformer along
with the windings parameters (Figure A.2), and Table A.2 presents its respective values.
Figure A.4 Transformer core geometry parameters.
Table A.2 Transformer geometry parameters.
Parameter Dimension (mm)

W 150.00
w’ 56.00
w’’ 25.40
Core
H 125.33
h’ 25.07
h’’ 25.40
D 62.00
WH 70.00
Windings
DL 8.00
DH 10.00
Appendix B
Thermal Parameters and Formulation
Thermal Parameters and Formulation | 87
B. Thermal Parameters and Formulation

This appendix details the thermal properties of the materials used on the analytical and
numerical thermal model of the transformer. Table B.1 summarizes the materials thermal
and physical properties used in the models.
Table B.1 Materials thermal and physical properties. [4]
Thermal conductivity k Emissivity

Material (W/°C/m) ε
Electrical steel sheet (lamination direction) 22-40
Electrical steel sheet (direction normal to
0.6
lamination)
Treating varnish 0.26
Black paint 0.9-0.95
Cast iron 0.3
The model used to calculate the thermal conductivity of the windings was shown
in section 3.4.4, Table B.2 shows the parameters values used to obtain the thermal
conductivity for each winding, the insulation thermal conductivity corresponds to a
treating varnish typically used in electric machines [4].
Table B.2 Windings thermal conductivities parameters.
ki (W/m/°C) d’ (mm) d (mm) δi (mm) kav

(W/m/°C)
High voltage
1.67 1.6 0.07 5.95
winding
0.26
Low voltage
2.074 2 0.074 7.04
winding
The convection coefficients were formulated by the relations found in [47] for free
air convection, the relations used are presented in Table B.3.
88 | Thermal Parameters and Formulation
Table B.3 Convection coefficients relations. (Adapted from [47].)
Geometry Convection coefficient (W/m2/°C)

Vertical plate or cylinder
Horizontal plate ( , being A the
surface area and p the perimeter)
(a) Hot surface facing up
(b) Hot surface facing down
By using the presented convection relations and the geometry parameters of the
transformer presented in Figure A.4 and Figure A.2, the convection coefficients are
formulated and shown in Table B.4.
Table B.4 Convection coefficients equations.
Convection coefficient (W/m2/°C)
Transformer’s Thermal Resistance Network Network

The TRN developed in section 3.4 was implemented on the circuit simulator Multisim.
Figure B.1 shows the implemented TRN. Table B.5 shows the temperature results on the
middle point of each element, and the temperature result for the surfaces in contact with
the ambient. It is important to notice, that these results correspond to the TRN
considering the transformer’s losses at ambient temperature.
Figure B.1 Developed TRN considering the transformer's losses at ambient temperature
Table B.5 TRN temperature results considering the transformer's losses at ambient temperature
Middle Point Surface xy Surface xz Surface yz

Volume
Temperature Temperature Temperature Temperature
Element
(°C) (°C) (°C) (°C)
Vw1 103 102 99.7 101
Vw2 103 102 101 101
Vw3 107 - - -
Vw4 107 - - -
Vw5 103 - 100 -
Vw6 103 - 102 -
Vw7 105 - - -
Vw8 106 - - -
Vc1 102 - - -
Vc2 94.7 84.4 94.1 -
Vc3 91.7 81.8 91.5 -
Vc4 89.4 79.8 89.2 89.2
Vc5 87.3 70.9 - 87
Vc6 91.3 82.7 - 91
Vc7 93.8 83.6 - -
Vc8 96.7 86.1 - -
Vc9 103 - - -
Vc10 103 - - -
Vc11 99.1 - 98.4 -
Vc12 97.2 - 96.9 -
Vc13 96.7 - 95.4 95.4
Vc14 94.7 - - 94.4
Vc15 97.5 - - 97.2
Vc16 99.3 - - -
Vc17 101 - - -
Vc18 104 - - -
Figure B.2 shows the implemented TRN considering the transformer’s losses at the
steady-state temperature, and Table B.6 shows the temperature results on the middle
point of each element, and the temperature result for the surfaces in contact with the
ambient.
Figure B.2 Developed TRN considering the transformer's losses at the steady-state temperature
Table B.6 TRN temperature results considering the transformer's losses at the steady-state temperature
Middle Point Surface xy Surface xz Surface yz

Volume
Temperature Temperature Temperature Temperature
Element
(°C) (°C) (°C) (°C)
Vw1 116 115 112 113
Vw2 117 116 115 113
Vw3 119 - - -
Vw4 120 - - -
Vw5 116 - 113 -
Vw6 118 - 116 -
Vw7 117 - - -
Vw8 117 - - -
Vc1 113 - - -
Vc2 104 92.3 104 -
Vc3 101 89.2 100 -
Vc4 98 86.9 97.6 97.6
Vc5 95.5 76.7 - 95.2
Vc6 100 93.1 - 99.8
Vc7 103 91.3 - -
Vc8 106 94.1 - -
Vc9 114 - - -
Vc10 115 - - -
Vc11 109 - 109 -
Vc12 107 - 107 -
Vc13 105 - 105 105
Vc14 104 - - 103
Vc15 107 - - 109
Vc16 109 - - -
Vc17 111 - - -
Vc18 115 - - -

Coupled Electromagneti GabrielMendes

Uploaded by

Copyright:

Available Formats

Coupled Electromagneti GabrielMendes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Coupled Electromagneti GabrielMendes

Uploaded by

Copyright:

Available Formats

Coupled Electromagnetic and Thermal

Analysis of Electric Machines

Gabriel Dias Mendes

Gabriel Dias Mendes

Dissertation presented to the School of Technology and Management of

Keywords: Finite Element Analysis, Thermal Model, Electromagnetic Model, Coupled

Palavras-chave: Análise por Elementos Finitos, Modelo Térmico, Modelo

List of Figures ........................................................................................................ xiii

Temperature gradient, °C/m or K/m

Rcond Conduction thermal resistance, °C/W or K/W

α Resistivity temperature coefficient, or

β Steinmetz’s equation parameter

1.2 Electric Machines Design and Modelling

1.4 Structure of the Work

2 Heat Generation and Transfer

2.1 Heat Generation

2.1.1 Resistive Loss

Frequency (Hz) 10 50 100 500 104 108 1010

Considering the skin effect, the resistive loss is given by

temperature rise of 55 °C in a copper wire, will result in a 20.96% increase of the

2.1.2 Iron Loss

2.2 Heat Transfer

Not all cases need to be treated as a 3D heat conduction problem, it is possible to

Figure 2.1 One-dimensional heat transfer by conduction (Adapted from [46])

Independently of the type of convection, the heat transfer rate by convection is

Process h [W/ m²/°C]

where σSB is the Stefan-Boltzmann constant that is equal to W/m2/K4, As is

being Tsur the surrounding surface temperature [47].

Figure 2.4 Radiation heat transfer (Adapted from [47]).

Figure 2.5 Directional distribution of emissivity (Adapted from [47]).

3 Analytical Thermal Modelling

3.1 Thermal Resistance Network

problem domain. To perform an appropriate TRN it is necessary to evaluate what level of

3.2 Thermal Resistance Network of Electric Machines

3.2.1 Heat Generation

3.3 Thermal Resistances

Where is the temperature difference, and the heat transfer rate.

Figure 3.1 Rectangle bar

By using equation the thermal resistance for conduction, or conduction resistance,

Figure 3.2 Hollow cylinder

The conduction resistance of the winding is a complex challenge, due to its

following equation is an example of a relation available in [47],

Figure 3.3 Convection surface

Although the equation is very simple, it addresses the temperature dependency of

being ε the surface emissivity, the Stefan-Boltzmann constant, Ts the surface

temperature and Tsur the surrounding temperature.

3.3.4 Volume Element

Figure 3.4 Volume element (Adapted from [56])

Obtaining a TRN of a 3D geometry can be easily developed recurring to the volume

3.4 Analytical Thermal Modelling of a Low Power

3.4.1 Transformer Geometry

3.4.2 Transformer Losses

3.4.3 Transformer Volume Element Modelling

winding elements, respectively.

Figure 3.7 Convection coefficients

conductivity of the core in the direction specified by the correspondent subscript.

Figure 3.8 Vc4 thermal resistances

3.4.4 Winding Thermal Conductivity

Figure 3.9 Winding structure (Adapted from [4])

The average conductivity for this type of winding is given by